- From: William A. Rowe Jr. <wrowe@rowe-clan.net>
- Date: Mon, 04 Oct 2010 11:47:55 -0500
- To: Julian Reschke <julian.reschke@gmx.de>
- CC: Adam Barth <w3c@adambarth.com>, Mark Nottingham <mnot@mnot.net>, HTTP Working Group <ietf-http-wg@w3.org>
On 10/4/2010 11:12 AM, Julian Reschke wrote: > On 03.10.2010 20:43, Julian Reschke wrote: >> ... >>> Do you have evidence that, for example, the percent encoding is rarely >>> used? In the absence of evidence, it's unlikely implementations will >>> remove support for the behavior. >> >> The only reason why legacy servers would use percent-escaping is because >> they were written for IE only. >> >> Every other server will sniff; and at least for the one case where I had >> to do this, I enabled the IE behavior only for IE. (Every other UA will >> get RFC2231 encoding). >> >> I can't provide empirical data, though. >> ... > > ...but I *can* provide a link to a thread where the same solution was mentioned: > > "...The solution at the end was to detect IE via the user agent header, do IE's way for > the filename and do the RFC 2231 way for others. But, > that's assumes UTF-8 URL encoding is turned on in IE..." -- > <http://lists.w3.org/Archives/Public/public-html/2008Mar/0119.html> In lieu of handling %-escapes, which were 1) semantically nonsensical in http response headers (while 2231 would be valid) yet 2) not fully nonconforming, since they were 7 bit ascii clean, so they did not trip up 7 bit handling... Why not simply drop filename= as a legacy, nonstandard representation, similar to the original cookie spec, and replace with name= or similar for all RFC2231 conforming names? This would allow both user agents and servers to continue to interoperate with legacy schemas, while offering a bridge to HTTP/1.1 semantics. The definition of a name= argument could reasonably be declared to override any filename= argument observed. Possibility?
Received on Monday, 4 October 2010 16:48:44 UTC