To finish(?) this thread on a positive note ... I've finally managed to beat our proxy from MarkLogic. Actually, using the "?" in between proxy and remote URL doesn't always work. What seems to always work however in our case, is to not use the question mark, but to only add the remote URL's path component, and to always set the Host HTTP header. Example for archival purposes below.
let $proxy := "http://my.proxy.org:80"; as xs:string; (: simple http proxy :) let $uri := "http://api.scribd.com/api"; as xs:string; (: target web service :) let $host := tokenize($uri,'/')[3] (: ==> api.scribd.com:) let $proxyuri := concat($proxy, '/', tokenize($uri, '/')[last()]) (: ==> http://my.proxy.org:80/api :) (: ... :) return xdmp:http-post( $proxyuri, <options xmlns="xdmp:http"> <headers> <Host>{$host}</Host> <Content-Type>multipart/form-data; boundary={$boundary}</Content-Type> </headers> </options>, $patched-multipart-encode ) cheers, Jakob. On Mon, Sep 26, 2011 at 16:23, Jakob Fix <[email protected]> wrote: > Oops, Geert was right, adding "?" between proxy and target url works. > > Now I only have to make http-post work, too! > > cheers, > Jakob. > > > > On Mon, Sep 26, 2011 at 16:16, Jakob Fix <[email protected]> wrote: >> I'd love if this worked for us, but it doesn't. We have a "websense >> proxy"[1]. The following is the output from Jim's query, please note >> the document returned corresponds to Google's 404 (which is quite >> geeky "!!1", horribly invalid, but oh so very much optimized) because >> somehow the proxy sends as the requested URL >> 'http://www.google.com/www.google.com/' which doesn't exist, >> apparently ... >> >> <response xmlns="xdmp:http"> >> <code>404</code> >> <message>Not Found</message> >> <headers> >> <content-type>text/html; charset=UTF-8</content-type> >> <x-content-type-options>nosniff</x-content-type-options> >> <date>Mon, 26 Sep 2011 14:02:19 GMT</date> >> <server>sffe</server> >> <x-xss-protection>1; mode=block</x-xss-protection> >> <x-cache>MISS from www-proxy.atmyorganisation.org</x-cache> >> <via>1.0 net-gw2.atmyorganisation.org:3128 (squid/2.7.STABLE7)</via> >> <content-length>11803</content-length> >> <age>0</age> >> </headers> >> </response> >> <!DOCTYPE html> >> <html lang=en> >> <meta charset=utf-8> >> <title>Error 404 (Not Found)!!1</title> >> <style> >> *{margin:0;padding:0}html,code{font:15px/22px >> arial,sans-serif}html{background:#fff;color:#222;padding:15px}body{background:url(data:image/png;base64,iVBORw0KGgoAAA >> [more base64 data cut]) 100% 5px no-repeat;margin:7% auto >> 0;max-width:390px;min-height:180px;padding:30px 0 15px}* > >> body{padding-right:205px}p{margin:22px 0 >> 0;overflow:hidden}ins,#g{text-decoration:none}ins{color:#777}a >> img{border:0}#g{background:url(data:image/png;base64,iVBORw0KGgoAAA[more >> base64 data cut]);display:block;height:55px;margin:0 0 >> -7px;width:150px}* > #g{margin-left:-2px}#g img{visibility:hidden}* >> html #g img{visibility:visible}*+html #g img{visibility:visible} >> </style> >> <a href=//www.google.com/ id=g><img >> src=//www.google.com/images/logo_sm.gif alt=Google></a> >> <p><b>404.</b> <ins>That’s an error.</ins> >> <p>The requested URL <code>/www.google.com/</code> was not found on >> this server. <ins>That’s all we know.</ins> >> >> >> so it looks like our proxy is mishandling the request url. Btw, just >> trying to use the url without the protocol (http://) doesn't work >> either. Any more ideas welcome, of course. >> >> cheers, >> Jakob. >> >> [1] http://en.wikipedia.org/wiki/Websense >> >> >> >> On Thu, Sep 22, 2011 at 13:14, James Fuller <[email protected]> >> wrote: >>> in the meantime you can do something like this; >>> >>> declare function local:http-get-proxy($proxy, $uri){ >>> let $host := fn:tokenize($uri,'/')[3] >>> let $proxyuri := fn:concat($proxy,$uri) >>> return >>> xdmp:http-get($proxyuri, >>> <options xmlns="xdmp:http"> <headers> <Host>{$host}</Host> >>> </headers></options> >>> ) >>> }; >>> >>> local:http-get-proxy('http://some.proxy.com:8080','http://www.google.com') >>> >>> works with all proxy setups I tested. >>> >>> gl, Jim Fuller >>> >>> On Thu, Sep 22, 2011 at 10:43 AM, Geert Josten <[email protected]> >>> wrote: >>>> I wouldn't mind it being a general configuration option at host level >>>> either, by the way. That should work for most cases I guess.. >>>> >>>> Kind regards, >>>> Geert >>>> >>>> -----Oorspronkelijk bericht----- >>>> Van: [email protected] >>>> [mailto:[email protected]] Namens Jakob Fix >>>> Verzonden: donderdag 22 september 2011 10:32 >>>> Aan: General MarkLogic Developer Discussion >>>> Onderwerp: Re: [MarkLogic Dev General] Will http-get support proxies in ML >>>> 5.0 ? >>>> >>>> Hear hear! +1 >>>> >>>> I had discussion with product management people at mluc this year >>>> where I requested this feature, but haven't had any news since, so am >>>> definitely interested also! >>>> >>>> Jakob. >>>> >>>> >>>> >>>> On Thu, Sep 22, 2011 at 09:34, Geert Josten <[email protected]> >>>> wrote: >>>>> Hi, >>>>> >>>>> I am once more plagued by running MarkLogic Server behind a proxy and >>>>> trying to use xdmp:http-get and related functions. Will MarkLogic Server >>>>> 5.0 support proxy options (server, port, user, pass)? >>>>> >>>>> Kind regards, >>>>> Geert >>>>> >>>>> _______________________________________________ >>>>> General mailing list >>>>> [email protected] >>>>> http://developer.marklogic.com/mailman/listinfo/general >>>>> >>>> _______________________________________________ >>>> General mailing list >>>> [email protected] >>>> http://developer.marklogic.com/mailman/listinfo/general >>>> _______________________________________________ >>>> General mailing list >>>> [email protected] >>>> http://developer.marklogic.com/mailman/listinfo/general >>>> >>> _______________________________________________ >>> General mailing list >>> [email protected] >>> http://developer.marklogic.com/mailman/listinfo/general >>> >> > _______________________________________________ General mailing list [email protected] http://developer.marklogic.com/mailman/listinfo/general
