From: Robin Berjon <robin@berjon.com>
To: Travis Leithead <travis.leithead@microsoft.com>
CC: Adrian Bateman <adrianba@microsoft.com>
Subject: Re: ReSpec patch for IE9+
Thread-Topic: ReSpec patch for IE9+
Thread-Index: Acyq71rnayhlIBE0SRKUSXAagJ7wqQDC6L8AAAH71EAAN/ArgA==
Date: Tue, 29 Nov 2011 14:03:09 +0000
Message-ID: <D623ED09-CB62-4F5B-B918-AF102A3A7215@berjon.com>
References: <9768D477C67135458BF978A45BCF9B38381BB8F6@TK5EX14MBXW602.wingroup.windeploy.ntdev.microsoft.com>
 <8A2004E9-AB57-4DF2-8AC4-D8A8C240F918@berjon.com>
 <9768D477C67135458BF978A45BCF9B38381BC3EC@TK5EX14MBXW602.wingroup.windeploy.ntdev.microsoft.com>
In-Reply-To: <9768D477C67135458BF978A45BCF9B38381BC3EC@TK5EX14MBXW602.wingroup.windeploy.ntdev.microsoft.com>
Content-Language: en-US
Content-Type: text/plain; charset="Windows-1252"
Content-ID: <741F5055B72D0E4D995CBBFED1335FE7@microsoft.com>
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0

On Nov 28, 2011, at 20:15 , Travis Leithead wrote:
> I found a couple more issues that I'd like to bring to your attention. Th=
ese aren't blockers for me, but I figured you'd like to know.
>
> 1)
>
> In the function "sh_getXMLHttpRequest" of "sh_main.min.js" (gnombat@users=
.sourceforge.net), there's a poor feature detect that routes IE down a lega=
cy ActiveX code path instead of using its native XMLHttpRequest object:

Yeech, this is horrible indeed, hadn't seen it. In general I'm not happy wi=
th the syntax highlighter in v1. It was the best thing available that I cou=
ld easily integrate at the time, but there are better options now. In v2 I =
use another one that has fewer problems =97 I'll see how hard it is to incl=
ude it in v1 as well (but probably not in the next two weeks).

> Specs that are hosted on sites other than "dev.w3.org" run into a cross-o=
rigin XHR problem when spec authors use <pre> tags with classes prefixed wi=
th "sh_". This happens because "sh_highlightDocument" is passed "this.base"=
 which is "dev.w3.org". Then, in the intervening logic of the minified code=
, any "sh_" prefixed class values end up using XHR to dynamically load the =
associated file. Of course, XHR's permissions use the document.domain's dom=
ain, and in some cases (Contacts API being one example; http://w3c-test.org=
/dap/contacts/) "dev.w3.org" is clearly cross-domain from "w3c-test.org".

Yes, I'm aware of this issue. Again, this is something that v2 fixes by add=
ing a build step that packages all the dependencies into one single file us=
ing RequireJS. Unfortunately, the contract that was funding my work on ReSp=
ec ended about a year ago so plans for v2 have been delayed quite a bit. W3=
C is looking into funding that since so many specs are now using it, so I h=
ave good hopes that I'll be able to take it out of alpha next year. Most of=
 it is already done (I use it for my own stuff), it mostly needs polish and=
 testing. Fingers crossed!

In the meantime I'll try to swap the sh_* stuff out as that's a useful shor=
t term improvement.

> I can live with the current results, but I figured you'd like to at least=
 be aware of this.

Thanks for the heads up!

One thing I forgot to mention in my previous email: don't hesitate to patch=
 and commit to ReSpec directly. We do commit then review (I get emails with=
 all the commits, and read them). It's a bit of a pain to patch since it's =
a good example of what happens when you start with a fifty lines hack and g=
row organically from there, but so far we've managed ;-)

--
Robin Berjon - http://berjon.com/ - @robinberjon