4 files changed, 388 insertions, 309 deletions

diff --git a/lib/Makefile.am b/lib/Makefile.am
index b278faf..185411f 100644
--- a/lib/Makefile.am
+++ b/lib/Makefile.am
@@ -1,30 +1,30 @@
 lib_LTLIBRARIES = libopkele.la
 
 DEFAULT_INCLUDES = -I${top_builddir}
 INCLUDES = \
            -I${top_srcdir}/include/ \
            ${KONFORKA_CFLAGS} \
            ${OPENSSL_CFLAGS} \
            ${LIBCURL_CPPFLAGS} \
            ${PCRE_CFLAGS} ${EXPAT_CFLAGS}
 libopkele_la_LIBADD = \
         ${LIBCURL} \
         ${PCRE_LIBS} ${EXPAT_LIBS} \
         ${OPENSSL_LIBS} \
         ${KONFORKA_LIBS}
 
 libopkele_la_SOURCES = \
                        params.cc \
                        util.cc \
                        server.cc \
                        secret.cc \
                        data.cc \
                        consumer.cc \
                        exception.cc \
                        extension.cc \
                        sreg.cc \
                        extension_chain.cc \
                        curl.cc expat.cc \
-                       openid_service_resolver.cc
+                       discovery.cc
 libopkele_la_LDFLAGS = \
         -version-info 2:0:0
diff --git a/lib/discovery.cc b/lib/discovery.cc
new file mode 100644
index 0000000..a35ce32
--- a/dev/null
+++ b/lib/discovery.cc
@@ -0,0 +1,375 @@
+#include <iostream>
+using namespace std;
+#include <list>
+#include <opkele/curl.h>
+#include <opkele/expat.h>
+#include <opkele/uris.h>
+#include <opkele/discovery.h>
+#include <opkele/exception.h>
+#include <opkele/util.h>
+
+#include "config.h"
+
+#define XRDS_HEADER "X-XRDS-Location"
+#define CT_HEADER "Content-Type"
+
+namespace opkele {
+    using std::list;
+    using xrd::XRD_t;
+    using xrd::service_t;
+
+    static const char *whitespace = " \t\r\n";
+    static const char *i_leaders = "=@+$!(";
+
+    static inline bool is_qelement(const XML_Char *n,const char *qen) {
+        return !strcasecmp(n,qen);
+    }
+    static inline bool is_element(const XML_Char *n,const char *en) {
+        if(!strcasecmp(n,en)) return true;
+        int nl = strlen(n), enl = strlen(en);
+        if( (nl>=(enl+1)) && n[nl-enl-1]=='\t'
+                && !strcasecmp(&n[nl-enl],en) )
+            return true;
+        return false;
+    }
+
+    static long element_priority(const XML_Char **a) {
+        for(;*a;++a)
+            if(!strcasecmp(*(a++),"priority")) {
+                long rv;
+                return (sscanf(*a,"%ld",&rv)==1)?rv:-1;
+            }
+        return -1;
+    }
+
+    class idigger_t : public util::curl_t, public util::expat_t {
+        public:
+            string xri_proxy;
+
+            enum {
+                xmode_html = 1, xmode_xrd = 2
+            };
+            int xmode;
+
+            string xrds_location;
+            string http_content_type;
+            service_t html_openid1;
+            service_t html_openid2;
+            string cdata_buf;
+            long status_code;
+            string status_string;
+
+            typedef list<string> pt_stack_t;
+            pt_stack_t pt_stack;
+            int skipping;
+
+            XRD_t *xrd;
+            service_t *xrd_service;
+            string* cdata;
+
+            idigger_t()
+                : util::curl_t(easy_init()),
+                util::expat_t(0),
+                xri_proxy(XRI_PROXY_URL) {
+                    CURLcode r;
+                    (r=misc_sets())
+                        || (r=set_write())
+                        || (r=set_header())
+                        ;
+                    if(r)
+                        throw exception_curl(OPKELE_CP_ "failed to set curly options",r);
+                }
+            ~idigger_t() throw() { }
+
+            void discover(idiscovery_t& result,const string& identity) {
+                result.clear();
+                string::size_type fsc = identity.find_first_not_of(whitespace);
+                if(fsc==string::npos)
+                    throw bad_input(OPKELE_CP_ "whtiespace-only identity");
+                string::size_type lsc = identity.find_last_not_of(whitespace);
+                assert(lsc!=string::npos);
+                if(!strncasecmp(identity.c_str()+fsc,"xri://",sizeof("xri://")-1))
+                    fsc += sizeof("xri://")-1;
+                if((fsc+1)>=lsc)
+                    throw bad_input(OPKELE_CP_ "not a character of importance in identity");
+                string id(identity,fsc,lsc-fsc+1);
+                if(strchr(i_leaders,id[0])) {
+                    result.normalized_id = id;
+                    /* TODO: further canonicalize xri identity? Like folding case  or whatever... */
+                    discover_at(
+                            result,
+                            xri_proxy + util::url_encode(id)+
+                            "?_xrd_r=application/xrd+xml;sep=false", xmode_xrd);
+                    if(status_code!=100)
+                        throw failed_xri_resolution(OPKELE_CP_
+                                "XRI resolution failed with '"+status_string+"' message",status_code);
+                    if(result.xrd.canonical_ids.empty())
+                        throw opkele::failed_discovery(OPKELE_CP_ "No CanonicalID for XRI identity found");
+                }else{
+                    if(id.find("://")==string::npos)
+                        id.insert(0,"http://");
+                    string::size_type fp = id.find('#');
+                    if(fp!=string::npos) {
+                        string::size_type qp = id.find('?');
+                        if(qp==string::npos || qp<fp)
+                            id.erase(fp);
+                        else if(qp>fp)
+                            id.erase(fp,qp-fp);
+                    }
+                    result.normalized_id = util::rfc_3986_normalize_uri(id);
+                    discover_at(result,id,xmode_html|xmode_xrd);
+                    const char * eu = 0;
+                    CURLcode r = easy_getinfo(CURLINFO_EFFECTIVE_URL,&eu);
+                    if(r)
+                        throw exception_curl(OPKELE_CP_ "failed to get CURLINFO_EFFECTIVE_URL",r);
+                    result.canonicalized_id = util::rfc_3986_normalize_uri(eu); /* XXX: strip fragment part? */
+                    if(xrds_location.empty()) {
+                        html2xrd(result.xrd);
+                    }else{
+                        discover_at(result,xrds_location,xmode_xrd);
+                        if(result.xrd.empty())
+                            html2xrd(result.xrd);
+                    }
+                }
+            }
+
+            void discover_at(idiscovery_t& result,const string& url,int xm) {
+                CURLcode r = easy_setopt(CURLOPT_URL,url.c_str());
+                if(r)
+                    throw exception_curl(OPKELE_CP_ "failed to set culry urlie",r);
+
+                (*(expat_t*)this) = parser_create_ns();
+                set_user_data(); set_element_handler();
+                set_character_data_handler();
+
+                xrds_location.clear(); http_content_type.clear();
+                xmode = xm;
+                if(xmode&xmode_html) {
+                    xrds_location.clear();
+                    html_openid1.clear(); html_openid2.clear();
+                }
+                xrd = &result.xrd;
+                cdata = 0; xrd_service = 0; skipping = 0;
+                status_code = 100; status_string.clear();
+
+                r = easy_perform();
+                if(r && r!=CURLE_WRITE_ERROR)
+                    throw exception_curl(OPKELE_CP_ "failed to perform curly request",r);
+
+                parse(0,0,true);
+            }
+
+            void html2xrd(XRD_t& x) {
+                if(!html_openid1.uris.empty()) {
+                    html_openid1.types.insert(STURI_OPENID11);
+                    x.services.add(-1,html_openid1);
+                }
+                if(!html_openid2.uris.empty()) {
+                    html_openid2.types.insert(STURI_OPENID20);
+                    x.services.add(-1,html_openid2);
+                }
+            }
+
+            size_t write(void *p,size_t s,size_t nm) {
+                if(skipping<0) return 0;
+                /* TODO: limit total size */
+                size_t bytes = s*nm;
+                parse((const char *)p,bytes,false);
+                return bytes;
+            }
+            size_t header(void *p,size_t s,size_t nm) {
+                size_t bytes = s*nm;
+                const char *h = (const char*)p;
+                const char *colon = (const char*)memchr(p,':',bytes);
+                const char *space = (const char*)memchr(p,' ',bytes);
+                if(space && ( (!colon) || space<colon ) ) {
+                    xrds_location.clear(); http_content_type.clear();
+                }else if(colon) {
+                    const char *hv = ++colon;
+                    int hnl = colon-h;
+                    int rb;
+                    for(rb = bytes-hnl-1;rb>0 && isspace(*hv);++hv,--rb);
+                    while(rb>0 && isspace(hv[rb-1])) --rb;
+                    if(rb) {
+                        if( (hnl>=sizeof(XRDS_HEADER))
+                                && !strncasecmp(h,XRDS_HEADER":",
+                                    sizeof(XRDS_HEADER)) ) {
+                            xrds_location.assign(hv,rb);
+                        }else if( (hnl>=sizeof(CT_HEADER))
+                                && !strncasecmp(h,CT_HEADER":",
+                                    sizeof(CT_HEADER)) ) {
+                            const char *sc = (const char*)memchr(
+                                    hv,';',rb);
+                            http_content_type.assign(hv,sc?(sc-hv):rb);
+                        }
+                    }
+                }
+                return curl_t::header(p,s,nm);
+            }
+
+            void start_element(const XML_Char *n,const XML_Char **a) {
+                if(skipping<0) return;
+                if(skipping) {
+                    if(xmode&xmode_html)
+                        html_start_element(n,a);
+                    ++skipping; return;
+                }
+                if(pt_stack.empty()) {
+                    if(is_qelement(n,NSURI_XRDS "\tXRDS"))
+                        return;
+                    if(is_qelement(n,NSURI_XRD "\tXRD")) {
+                        assert(xrd);
+                        xrd->clear();
+                        pt_stack.push_back(n);
+                    }else if(xmode&xmode_html) {
+                        html_start_element(n,a);
+                    }else{
+                        skipping = -1;
+                    }
+                }else{
+                    int pt_s = pt_stack.size();
+                    if(pt_s==1) {
+                        /* TODO: xrd:XRD/xrd:Expires */
+                        if(is_qelement(n,NSURI_XRD "\tCanonicalID")) {
+                            assert(xrd);
+                            cdata = &(xrd->canonical_ids.add(element_priority(a),string()));
+                        }else if(is_qelement(n,NSURI_XRD "\tLocalID")) {
+                            assert(xrd);
+                            cdata = &(xrd->local_ids.add(element_priority(a),string()));
+                        }else if(is_qelement(n,NSURI_XRD "\tService")) {
+                            assert(xrd);
+                            xrd_service = &(xrd->services.add(element_priority(a),
+                                        service_t()));
+                            pt_stack.push_back(n);
+                        }else if(is_qelement(n,NSURI_XRD "\tStatus")) {
+                            for(;*a;) {
+                                if(!strcasecmp(*(a++),"code")) {
+                                    if(sscanf(*(a++),"%ld",&status_code)==1 && status_code!=100) {
+                                        cdata = &status_string;
+                                        pt_stack.push_back(n);
+                                        break;
+                                    }
+                                }
+                            }
+                        }else if(xmode&xmode_html) {
+                            html_start_element(n,a);
+                        }else{
+                            skipping = 1;
+                        }
+                    }else if(pt_s==2) {
+                        if(is_qelement(pt_stack.back().c_str(), NSURI_XRD "\tService")) {
+                            if(is_qelement(n,NSURI_XRD "\tType")) {
+                                assert(xrd); assert(xrd_service);
+                                cdata_buf.clear();
+                                cdata = &cdata_buf;
+                            }else if(is_qelement(n,NSURI_XRD "\tURI")) {
+                                assert(xrd); assert(xrd_service);
+                                cdata = &(xrd_service->uris.add(element_priority(a),string()));
+                            }else if(is_qelement(n,NSURI_XRD "\tLocalID")
+                                    || is_qelement(n,NSURI_OPENID10 "\tDelegate") ) {
+                                assert(xrd); assert(xrd_service);
+                                cdata = &(xrd_service->uris.add(element_priority(a),string()));
+                            }else{
+                                skipping = 1;
+                            }
+                        }else
+                            skipping = 1;
+                    }else if(xmode&xmode_html) {
+                        html_start_element(n,a);
+                    }else{
+                        skipping = 1;
+                    }
+                }
+            }
+            void end_element(const XML_Char *n) {
+                if(skipping<0) return;
+                if(skipping) {
+                    --skipping; return;
+                }
+                if(is_qelement(n,NSURI_XRD "\tType")) {
+                    assert(xrd); assert(xrd_service); assert(cdata==&cdata_buf);
+                    xrd_service->types.insert(cdata_buf);
+                }else if(is_qelement(n,NSURI_XRD "\tService")) {
+                    assert(xrd); assert(xrd_service);
+                    assert(!pt_stack.empty());
+                    assert(pt_stack.back()==(NSURI_XRD "\tService"));
+                    pt_stack.pop_back();
+                    xrd_service = 0;
+                }else if(is_qelement(n,NSURI_XRD "\tStatus")) {
+                    assert(xrd);
+                    if(is_qelement(pt_stack.back().c_str(),n)) {
+                        assert(cdata==&status_string);
+                        pt_stack.pop_back();
+                        if(status_code!=100)
+                            skipping = -1;
+                    }
+                }else if((xmode&xmode_html) && is_element(n,"head")) {
+                    skipping = -1;
+                }
+                cdata = 0;
+            }
+            void character_data(const XML_Char *s,int l) {
+                if(skipping) return;
+                if(cdata) cdata->append(s,l);
+            }
+
+            void html_start_element(const XML_Char *n,const XML_Char **a) {
+                if(is_element(n,"meta")) {
+                    bool heq = false;
+                    string l;
+                    for(;*a;a+=2) {
+                        if(!( strcasecmp(a[0],"http-equiv")
+                                || strcasecmp(a[1],XRDS_HEADER) ))
+                            heq = true;
+                        else if(!strcasecmp(a[0],"content"))
+                            l.assign(a[1]);
+                    }
+                    if(heq)
+                        xrds_location = l;
+                }else if(is_element(n,"link")) {
+                    string rels;
+                    string href;
+                    for(;*a;a+=2) {
+                        if( !strcasecmp(a[0],"rel") ) {
+                            rels.assign(a[1]);
+                        }else if( !strcasecmp(a[0],"href") ) {
+                            const char *ns = a[1];
+                            for(;*ns && isspace(*ns);++ns);
+                            href.assign(ns);
+                            string::size_type lns=href.find_last_not_of(whitespace);
+                            href.erase(lns+1);
+                        }
+                    }
+                    for(string::size_type ns=rels.find_first_not_of(whitespace);
+                            ns!=string::npos; ns=rels.find_first_not_of(whitespace,ns)) {
+                        string::size_type s = rels.find_first_of(whitespace,ns);
+                        string rel;
+                        if(s==string::npos) {
+                            rel.assign(rels,ns,string::npos);
+                            ns = string::npos;
+                        }else{
+                            rel.assign(rels,ns,s-ns);
+                            ns = s;
+                        }
+                        if(rel=="openid.server")
+                            html_openid1.uris.add(-1,href);
+                        else if(rel=="openid.delegate")
+                            html_openid1.local_ids.add(-1,href);
+                        else if(rel=="openid2.provider")
+                            html_openid2.uris.add(-1,href);
+                        else if(rel=="openid2.local_id")
+                            html_openid2.local_ids.add(-1,href);
+                    }
+                }else if(is_element(n,"body")) {
+                    skipping = -1;
+                }
+            }
+
+    };
+
+    void idiscover(idiscovery_t& result,const string& identity) {
+        idigger_t idigger;
+        idigger.discover(result,identity);
+    }
+
+}
diff --git a/lib/openid_service_resolver.cc b/lib/openid_service_resolver.cc
deleted file mode 100644
index 505e5b2..0000000
--- a/lib/openid_service_resolver.cc
+++ b/dev/null
@@ -1,298 +0,0 @@
-#include <cctype>
-#include <opkele/exception.h>
-#include <opkele/util.h>
-#include <opkele/openid_service_resolver.h>
-#include <opkele/uris.h>
-
-#define LOCATION_HEADER "X-XRDS-Location"
-
-namespace opkele {
-    static const char *whitespace = " \t\r\n";
-
-    openid_service_resolver_t::openid_service_resolver_t(const string& xp)
-        : util::curl_t(easy_init()),
-        util::expat_t(0),
-        xri_proxy(xp.empty()?"http://beta.xri.net/":xp)
-    {
-        CURLcode r;
-        (r=misc_sets())
-        || (r=set_write())
-        || (r==set_header())
-        ;
-        if(r)
-            throw opkele::exception_curl(OPKELE_CP_ "failed to set curly options",r);
-    }
-
-    static bool is_element(const XML_Char *n,const char *en) {
-        if(!strcasecmp(n,en)) return true;
-        int nl = strlen(n), enl = strlen(en);
-        if( (nl>=(enl+1)) && n[nl-enl-1]=='\t'
-                && !strcasecmp(&n[nl-enl],en) )
-            return true;
-        return false;
-    }
-    static inline bool is_qelement(const XML_Char *n,const char *qen) {
-        return !strcasecmp(n,qen);
-    }
-    static inline bool is_element(
-            const openid_service_resolver_t::parser_node_t& n,
-            const char *en) {
-        return is_element(n.element.c_str(),en);
-    }
-    static inline bool is_qelement(
-            const openid_service_resolver_t::parser_node_t& n,
-            const char *qen) {
-        return is_qelement(n.element.c_str(),qen);
-    }
-
-    void openid_service_resolver_t::start_element(const XML_Char *n,const XML_Char **a) {
-        if(state!=state_parse) return;
-        tree.push(n,a);
-        parser_node_t& t = tree.top();
-        if(is_element(n,"html") || is_element(n,"head")
-                || is_qelement(n,NSURI_XRDS "\tXRDS")
-                || is_qelement(n,NSURI_XRD "\tXRD") )
-            t.skip_tags = false;
-        else if(is_qelement(n,NSURI_XRD "\tService")
-                || is_qelement(n,NSURI_XRD "\tType")
-                || is_qelement(n,NSURI_XRD "\tURI")
-                || is_qelement(n,NSURI_OPENID10 "\tDelegate")
-                || is_qelement(n,NSURI_XRD "\tCanonicalID") )
-            t.skip_tags = t.skip_text = false;
-        else if(is_element(n,"body"))
-            state = state_stopping_body;
-    }
-    void openid_service_resolver_t::end_element(const XML_Char *n) {
-        if(state!=state_parse) return;
-        assert(tree.top().element == n);
-        pop_tag();
-    }
-    void openid_service_resolver_t::character_data(const XML_Char *s,int l) {
-        if(state!=state_parse) return;
-        if( !( tree.empty() || tree.top().skip_text ) )
-            tree.top().content.append(s,l);
-    }
-
-    static void copy_trim_whitespace(string& to,const string& from) {
-        string::size_type ns0 = from.find_first_not_of(whitespace);
-        if(ns0==string::npos) {
-            to.clear(); return;
-        }
-        string::size_type ns1 = from.find_last_not_of(whitespace);
-        assert(ns1!=string::npos);
-        to.assign(from,ns0,ns1-ns0+1);
-    }
-
-    void openid_service_resolver_t::pop_tag() {
-        assert(!tree.empty());
-        parser_node_t& t = tree.top();
-        if( is_element(t,"meta")
-                && !strcasecmp(t.attrs["http-equiv"].c_str(),LOCATION_HEADER) ) {
-            xrds_location = t.attrs["content"];
-        }else if( is_element(t,"link") ) {
-            parser_node_t::attrs_t::const_iterator ir = t.attrs.find("rel");
-            if(ir!=t.attrs.end()) {
-                const string& rels = ir->second;
-                for(string::size_type ns = rels.find_first_not_of(whitespace);
-                        ns!=string::npos;
-                        ns=rels.find_first_not_of(whitespace,ns)) {
-                    string::size_type s = rels.find_first_of(whitespace,ns);
-                    string rel;
-                    if(s==string::npos) {
-                        rel.assign(rels,ns,string::npos);
-                        ns = string::npos;
-                    }else{
-                        rel.assign(rels,ns,s-ns);
-                        ns = s;
-                    }
-                    if(rel=="openid.server")
-                        copy_trim_whitespace(html_SEP.xrd_URI,t.attrs["href"]);
-                    else if(rel=="openid.delegate")
-                        copy_trim_whitespace(html_SEP.openid_Delegate,t.attrs["href"]);
-                }
-            }
-        }else if( is_element(t,"head") )
-            state = state_stopping_head;
-        else if( is_qelement(t,NSURI_XRD "\tXRD")) {
-            if( !(
-                        (
-                         xri_mode
-                         && t.auth_info.canonical_id.empty()
-                        ) ||
-                        t.auth_info.auth_SEP.xrd_Type.empty()
-                 ) )
-                auth_info = t.auth_info;
-        }else if( tree.size()>1 ) {
-            parser_node_t& p = tree.parent();
-            if( is_qelement(p,NSURI_XRD "\tService") ) {
-                if( is_qelement(t,NSURI_XRD "\tType") ) {
-                    if(t.content==STURI_OPENID10) {
-                        string tmp; copy_trim_whitespace(tmp,t.content);
-                        p.auth_info.auth_SEP.xrd_Type.insert(tmp);
-                    }
-                }else if( is_qelement(t,NSURI_XRD "\tURI") )
-                    copy_trim_whitespace(p.auth_info.auth_SEP.xrd_URI,t.content);
-                else if( is_qelement(t,NSURI_OPENID10 "\tDelegate") )
-                    copy_trim_whitespace(p.auth_info.auth_SEP.openid_Delegate,t.content);
-            }else if( is_qelement(p,NSURI_XRD "\tXRD") ) {
-                if(is_qelement(t,NSURI_XRD "\tService") ) {
-                    if( !t.auth_info.auth_SEP.xrd_Type.empty() ) {
-                        parser_node_t::attrs_t::const_iterator ip
-                            = t.attrs.find("priority");
-                        if(ip!=t.attrs.end()) {
-                            const char *nptr = ip->second.c_str();
-                            char *eptr = 0;
-                            t.auth_info.auth_SEP.priority = strtol(nptr,&eptr,10);
-                            if(nptr==eptr)
-                                t.auth_info.auth_SEP.priority = LONG_MAX;
-                        }
-                        if( (t.auth_info.auth_SEP.priority < p.auth_info.auth_SEP.priority)
-                                || p.auth_info.auth_SEP.xrd_Type.empty() )
-                            p.auth_info.auth_SEP = t.auth_info.auth_SEP;
-                    }
-                }else if( is_qelement(t,NSURI_XRD "\tCanonicalID") )
-                    copy_trim_whitespace(p.auth_info.canonical_id,t.content);
-            }
-        }
-        
-        tree.pop();
-    }
-
-    size_t openid_service_resolver_t::write(void *p,size_t s,size_t nm) {
-        if(state != state_parse)
-            return 0;
-        /* TODO: limit total size */
-        size_t bytes = s*nm;
-        parse((const char *)p,bytes,false);
-        return bytes;
-    }
-
-    size_t openid_service_resolver_t::header(void *p,size_t s,size_t nm) {
-        size_t bytes = s*nm;
-        const char *h = (const char *)p;
-        const char *colon = (const char*)memchr(p,':',bytes);
-        const char *space = (const char*)memchr(p,' ',bytes);
-        if(space && ( (!colon) || space<colon ) ) {
-            xrds_location.clear(); http_content_type.clear();
-        }else if(colon) {
-            const char *hv = ++colon;
-            int hnl = colon-h;
-            int rb;
-            for(rb = bytes-hnl-1;
-                    rb>0 && isspace(*hv);
-                    ++hv,--rb );
-            while(rb>0 && isspace(hv[rb-1]))
-                --rb;
-            if(rb) {
-                if( (hnl >= sizeof(LOCATION_HEADER))
-                        && !strncasecmp(h,LOCATION_HEADER ":",
-                            sizeof(LOCATION_HEADER)) ) {
-                    xrds_location.assign(hv,rb);
-                }else if( (hnl >= sizeof("Content-Type"))
-                        && !strncasecmp(h,"Content-Type:",
-                            sizeof("Content-Type")) ) {
-                    const char *sc = (const char*)memchr(
-                            hv,';',rb);
-                    http_content_type.assign(
-                            hv,sc?(sc-hv):rb );
-                }
-            }
-        }
-        return curl_t::header(p,s,nm);
-    }
-
-    void openid_service_resolver_t::discover_service(const string& url,bool xri) {
-        CURLcode r = easy_setopt(CURLOPT_URL,url.c_str());
-        if(r)
-            throw opkele::exception_curl(OPKELE_CP_ "failed to set curly urlie",r);
-
-        (*(expat_t*)this) = parser_create_ns();
-        set_user_data(); set_element_handler();
-        set_character_data_handler();
-        tree.clear();
-        state = state_parse;
-
-        r = easy_perform();
-        if(r && r!=CURLE_WRITE_ERROR)
-            throw exception_curl(OPKELE_CP_ "failed to perform curly request",r);
-
-        parse(0,0,true);
-        while(!tree.empty()) pop_tag();
-    }
-
-    const openid_auth_info_t& openid_service_resolver_t::resolve(const string& id) {
-        auth_info = openid_auth_info_t();
-        html_SEP = openid_auth_SEP_t();
-
-        string::size_type fns = id.find_first_not_of(whitespace);
-        if(fns==string::npos)
-            throw opkele::bad_input(OPKELE_CP_ "whitespace-only identity");
-        string::size_type lns = id.find_last_not_of(whitespace);
-        assert(lns!=string::npos);
-        if(!strncasecmp(
-                    id.c_str()+fns,"xri://",
-                    sizeof("xri://")-1))
-            fns+=sizeof("xri://")-1;
-        string nid(id,fns,lns-fns+1);
-        if(nid.empty())
-            throw opkele::bad_input(OPKELE_CP_ "nothing significant in identity");
-        if(strchr("=@+$!(",*nid.c_str())) {
-            discover_service(
-                    xri_proxy + util::url_encode(nid) +
-                    "?_xrd_t=" STURI_OPENID10 "&_xrd_r=application/xrd+xml;sep=true",
-                    true );
-            if(auth_info.canonical_id.empty()
-                    || auth_info.auth_SEP.xrd_Type.empty() )
-                throw opkele::failed_lookup(OPKELE_CP_ "no OpenID service for XRI found");
-        }else{
-            const char *np = nid.c_str();
-            if( (strncasecmp(np,"http",4) || strncmp(
-                            tolower(*(np+4))=='s'? np+5 : np+4, "://", 3))
-#ifndef NDEBUG
-                    && strncasecmp(np,"file:///",sizeof("file:///")-1)
-        #endif  /* XXX: or how do I let tests work? */
-                    )
-                nid.insert(0,"http://");
-            string::size_type fp = nid.find('#');
-            if(fp!=string::npos) {
-                string::size_type qp = nid.find('?');
-                if(qp==string::npos || qp<fp) {
-                    nid.erase(fp);
-                }else if(qp>fp)
-                    nid.erase(fp,qp-fp);
-            }
-            discover_service(nid);
-            const char *eu = 0;
-            CURLcode r = easy_getinfo(CURLINFO_EFFECTIVE_URL,&eu);
-            if(r)
-                throw exception_curl(OPKELE_CP_ "failed to get CURLINFO_EFFECTIVE_URL",r);
-            string canonicalized_id = util::rfc_3986_normalize_uri(eu);
-            if(xrds_location.empty()) {
-                if(auth_info.auth_SEP.xrd_Type.empty()) {
-                    if(html_SEP.xrd_URI.empty())
-                        throw opkele::failed_lookup(OPKELE_CP_ "no OpenID 1.0 service discovered");
-                    auth_info.auth_SEP = html_SEP;
-                    auth_info.auth_SEP.xrd_Type.clear(); auth_info.auth_SEP.xrd_Type.insert( STURI_OPENID10 );
-                    auth_info.canonical_id = canonicalized_id;
-                }else{
-                    if(auth_info.canonical_id.empty())
-                        auth_info.canonical_id = canonicalized_id;
-                }
-            }else{
-                discover_service(xrds_location);
-                if(auth_info.auth_SEP.xrd_Type.empty()) {
-                    if(html_SEP.xrd_URI.empty())
-                        throw opkele::failed_lookup(OPKELE_CP_ "no OpenID 1.0 service discovered");
-                    auth_info.auth_SEP = html_SEP;
-                    auth_info.auth_SEP.xrd_Type.clear(); auth_info.auth_SEP.xrd_Type.insert( STURI_OPENID10 );
-                    auth_info.canonical_id = canonicalized_id;
-                }else{
-                    if(auth_info.canonical_id.empty())
-                        auth_info.canonical_id = canonicalized_id;
-                }
-            }
-        }
-        return auth_info;
-    }
-
-}
diff --git a/lib/util.cc b/lib/util.cc
index 69d37b5..83f0eef 100644
--- a/lib/util.cc
+++ b/lib/util.cc
@@ -1,303 +1,305 @@
 #include <errno.h>
 #include <cassert>
 #include <cctype>
 #include <cstring>
 #include <vector>
 #include <string>
 #include <stack>
 #include <openssl/bio.h>
 #include <openssl/evp.h>
 #include <curl/curl.h>
 #include "opkele/util.h"
 #include "opkele/exception.h"
 
 namespace opkele {
     using namespace std;
 
     namespace util {
 
         /*
          * base64
          */
         string encode_base64(const void *data,size_t length) {
             BIO *b64 = 0, *bmem = 0;
             try {
                 b64 = BIO_new(BIO_f_base64());
                 if(!b64)
                     throw exception_openssl(OPKELE_CP_ "failed to BIO_new() base64 encoder");
                 BIO_set_flags(b64,BIO_FLAGS_BASE64_NO_NL);
                 bmem = BIO_new(BIO_s_mem());
                 BIO_set_flags(b64,BIO_CLOSE);
                 if(!bmem)
                     throw exception_openssl(OPKELE_CP_ "failed to BIO_new() memory buffer");
                 BIO_push(b64,bmem);
                 if(((size_t)BIO_write(b64,data,length))!=length)
                     throw exception_openssl(OPKELE_CP_ "failed to BIO_write()");
                 if(BIO_flush(b64)!=1)
                     throw exception_openssl(OPKELE_CP_ "failed to BIO_flush()");
                 char *rvd;
                 long rvl = BIO_get_mem_data(bmem,&rvd);
                 string rv(rvd,rvl);
                 BIO_free_all(b64);
                 return rv;
             }catch(...) {
                 if(b64) BIO_free_all(b64);
                 throw;
             }
         }
 
         void decode_base64(const string& data,vector<unsigned char>& rv) {
             BIO *b64 = 0, *bmem = 0;
             rv.clear();
             try {
                 bmem = BIO_new_mem_buf((void*)data.data(),data.size());
                 if(!bmem)
                     throw exception_openssl(OPKELE_CP_ "failed to BIO_new_mem_buf()");
                 b64 = BIO_new(BIO_f_base64());
                 if(!b64)
                     throw exception_openssl(OPKELE_CP_ "failed to BIO_new() base64 decoder");
                 BIO_set_flags(b64,BIO_FLAGS_BASE64_NO_NL);
                 BIO_push(b64,bmem);
                 unsigned char tmp[512];
                 size_t rb = 0;
                 while((rb=BIO_read(b64,tmp,sizeof(tmp)))>0)
                     rv.insert(rv.end(),tmp,&tmp[rb]);
                 BIO_free_all(b64);
             }catch(...) {
                 if(b64) BIO_free_all(b64);
                 throw;
             }
         }
 
         /*
          * big numerics
          */
 
         BIGNUM *base64_to_bignum(const string& b64) {
             vector<unsigned char> bin;
             decode_base64(b64,bin);
             BIGNUM *rv = BN_bin2bn(&(bin.front()),bin.size(),0);
             if(!rv)
                 throw failed_conversion(OPKELE_CP_ "failed to BN_bin2bn()");
             return rv;
         }
 
         BIGNUM *dec_to_bignum(const string& dec) {
             BIGNUM *rv = 0;
             if(!BN_dec2bn(&rv,dec.c_str()))
                 throw failed_conversion(OPKELE_CP_ "failed to BN_dec2bn()");
             return rv;
         }
 
         string bignum_to_base64(const BIGNUM *bn) {
             vector<unsigned char> bin(BN_num_bytes(bn)+1);
             unsigned char *binptr = &(bin.front())+1;
             int l = BN_bn2bin(bn,binptr);
             if(l && (*binptr)&0x80){
                 (*(--binptr)) = 0; ++l;
             }
             return encode_base64(binptr,l);
         }
 
         /*
          * w3c times
          */
 
         string time_to_w3c(time_t t) {
             struct tm tm_t;
             if(!gmtime_r(&t,&tm_t))
                 throw failed_conversion(OPKELE_CP_ "failed to BN_dec2bn()");
             char rv[25];
             if(!strftime(rv,sizeof(rv)-1,"%Y-%m-%dT%H:%M:%SZ",&tm_t))
                 throw failed_conversion(OPKELE_CP_ "failed to strftime()");
             return rv;
         }
 
         time_t w3c_to_time(const string& w) {
             struct tm tm_t;
             memset(&tm_t,0,sizeof(tm_t));
             if(
                     sscanf(
                         w.c_str(),
                         "%04d-%02d-%02dT%02d:%02d:%02dZ",
                         &tm_t.tm_year,&tm_t.tm_mon,&tm_t.tm_mday,
                         &tm_t.tm_hour,&tm_t.tm_min,&tm_t.tm_sec
                     ) != 6 )
                 throw failed_conversion(OPKELE_CP_ "failed to sscanf()");
             tm_t.tm_mon--;
             tm_t.tm_year-=1900;
             time_t rv = mktime(&tm_t);
             if(rv==(time_t)-1)
                 throw failed_conversion(OPKELE_CP_ "failed to mktime()");
             return rv;
         }
 
         /*
          *
          */
 
         string url_encode(const string& str) {
             char * t = curl_escape(str.c_str(),str.length());
             if(!t)
                 throw failed_conversion(OPKELE_CP_ "failed to curl_escape()");
             string rv(t);
             curl_free(t);
             return rv;
         }
 
         string long_to_string(long l) {
             char rv[32];
             int r=snprintf(rv,sizeof(rv),"%ld",l);
             if(r<0 || r>=(int)sizeof(rv))
                 throw failed_conversion(OPKELE_CP_ "failed to snprintf()");
             return rv;
         }
 
         long string_to_long(const string& s) {
             char *endptr = 0;
             long rv = strtol(s.c_str(),&endptr,10);
             if((!endptr) || endptr==s.c_str())
                 throw failed_conversion(OPKELE_CP_ "failed to strtol()");
             return rv;
         }
 
         /*
          * Normalize URL according to the rules, described in rfc 3986, section 6
          *
          * - uppercase hext triplets (e.g. %ab -> %AB)
          * - lowercase scheme and host
          * - decode %-encoded characters, specified as unreserved in rfc 3986, section 2.3,
          *   that is - [:alpha:][:digit:]._~-
          * - remove dot segments
          * - remove empty and default ports
          * - if there's no path component, add '/'
          */
          string rfc_3986_normalize_uri(const string& uri) {
              static const char *whitespace = " \t\r\n";
              string rv;
              string::size_type ns = uri.find_first_not_of(whitespace);
              if(ns==string::npos)
                  throw bad_input(OPKELE_CP_ "Can't normalize empty URI");
              string::size_type colon = uri.find(':',ns);
              if(colon==string::npos)
                  throw bad_input(OPKELE_CP_ "No scheme specified in URI");
              transform(
                      uri.begin()+ns, uri.begin()+colon+1,
                      back_inserter(rv), ::tolower );
              bool s;
-             if(rv=="http:")
-                 s = false;
-             else if(rv=="https:")
-                 s = true;
-#ifndef NDEBUG
-             else if(rv=="file:")
-                 s = false;
-#endif /* XXX: or try to make tests work some other way */
-             else
-                 throw not_implemented(OPKELE_CP_ "Only http(s) URIs can be normalized here");
              string::size_type ul = uri.find_last_not_of(whitespace)+1;
              if(ul <= (colon+3))
                  throw bad_input(OPKELE_CP_ "Unexpected end of URI being normalized encountered");
              if(uri[colon+1]!='/' || uri[colon+2]!='/')
                  throw bad_input(OPKELE_CP_ "Unexpected input in URI being normalized after scheme component");
+             if(rv=="http:")
+                 s = false;
+             else if(rv=="https:")
+                 s = true;
+             else{
+                 /* TODO: support more schemes.
+                  * e.g. xri. How do we normalize
+                  * xri?
+                  */
+                 rv.append(uri,colon+1,ul-colon-1);
+                 return rv;
+             }
              rv += "//";
              string::size_type interesting = uri.find_first_of(":/#?",colon+3);
              if(interesting==string::npos) {
                  transform(
                          uri.begin()+colon+3,uri.begin()+ul,
                          back_inserter(rv), ::tolower );
                  rv += '/'; return rv;
              }
              transform(
                      uri.begin()+colon+3,uri.begin()+interesting,
                      back_inserter(rv), ::tolower );
              bool qf = false;
              char ic = uri[interesting];
              if(ic==':') {
                  string::size_type ni = uri.find_first_of("/#?%",interesting+1);
                  const char *nptr = uri.data()+interesting+1;
                  char *eptr = 0;
                  long port = strtol(nptr,&eptr,10);
                  if( (port>0) && (port<65535) && port!=(s?443:80) ) {
                      char tmp[6];
                      snprintf(tmp,sizeof(tmp),"%ld",port);
                      rv += ':'; rv += tmp;
                  }
                  if(ni==string::npos) {
                      rv += '/'; return rv;
                  }
                  interesting = ni;
              }else if(ic!='/') {
                  rv += '/'; rv += ic;
                  qf = true;
                  ++interesting;
              }
              string::size_type n = interesting;
              char tmp[3] = { 0,0,0 };
              stack<string::size_type> psegs; psegs.push(rv.length());
              string pseg;
              for(;n<ul;) {
                  string::size_type unsafe = uri.find_first_of(qf?"%":"%/?#",n);
                  if(unsafe==string::npos) {
                      pseg.append(uri,n,ul-n-1); n = ul-1;
                  }else{
                      pseg.append(uri,n,unsafe-n);
                      n = unsafe;
                  }
                  char c = uri[n++];
                  if(c=='%') {
                      if((n+1)>=ul)
                          throw bad_input(OPKELE_CP_ "Unexpected end of URI encountered while parsing percent-encoded character");
                      tmp[0] = uri[n++];
                      tmp[1] = uri[n++];
                      if(!( isxdigit(tmp[0]) && isxdigit(tmp[1]) ))
                          throw bad_input(OPKELE_CP_ "Invalid percent-encoded character in URI being normalized");
                      int cc = strtol(tmp,0,16);
                      if( isalpha(cc) || isdigit(cc) || strchr("._~-",cc) )
                          pseg += cc;
                      else{
                          pseg += '%';
                          pseg += toupper(tmp[0]); pseg += toupper(tmp[1]);
                      }
                  }else if(qf) {
                      rv += pseg; rv += c;
                      pseg.clear();
                  }else if(n>=ul || strchr("?/#",c)) {
                      if(pseg.empty() || pseg==".") {
                      }else if(pseg=="..") {
                          if(psegs.size()>1) {
                              rv.resize(psegs.top()); psegs.pop();
                          }
                      }else{
                          psegs.push(rv.length());
                          if(c!='/') {
                              pseg += c;
                              qf = true;
                          }
                          rv += '/'; rv += pseg;
                      }
                      if(c=='/' && (n>=ul || strchr("?#",uri[n])) ) {
                          rv += '/';
                          if(n<ul)
                              qf = true;
                      }else if(strchr("?#",c)) {
                          if(psegs.size()==1 && psegs.top()==rv.length())
                              rv += '/';
                          if(pseg.empty())
                              rv += c;
                          qf = true;
                      }
                      pseg.clear();
                  }else{
                      pseg += c;
                  }
              }
              if(!pseg.empty()) {
                  rv += '/'; rv += pseg;
              }
              return rv;
          }
 
     }
 
 }