httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Roy T. Fielding" <field...@kiwi.ICS.UCI.EDU>
Subject Re: MSIE + byteranges
Date Mon, 17 Feb 1997 00:27:10 GMT
Alexei Kosut writes:
>Agreed, but I disagree that they'll "say" we aren't speaking
>HTTP/1.1. They'll be compatible with HTTP/1.1 if they want to, and if
>they want to screw it up, history has shown that Apache's
>implementation of things has never had an affect on them.

Apache has far more pull now than it did before.  I see that more than
most because of my standards work, but there was a serious mindset
adjustment last Fall.  These people now depend on Apache for their
core business, because most of the good Web sites run Apache.

>It goes back to what I said earlier: Request-Range indicates support
>for draft-luotenen-byte-ranges-02 (or whatever the name was), which is
>seperate and distinct from HTTP/1.1, and Apache supports it.

I don't -- that draft had several distinct bugs that were fixed in HTTP/1.1,
and it is HTTP/1.1 that we support.

>You had no problem with this when I stuck it in last week; why is it
>that now you suddenly do, after I decided that we should check for
>MSIE's user-agent?

I believe in not hindering other people's business, but promoting progress.
Backwards-compatibility is only a good thing if it doesn't accumulate,
which is why a user-agent solution (with version numbers) is the best
way to account for browser bugs.  I didn't mention anything until you
offerred a choice as to methods, at which point I said which method
I preferred (just checking for "Mozilla/[23]" before sending byteranges).

I would also like to see an ifdef for STRICT_HTTP_COMPLIANCE; aside
from being an escape valve, it serves as a reminder for what needs to
be fixed in the spec and/or world.

>> The Mozilla clones can go suck an egg; they asked for bugwards-compatibility,
>> so that's what they get (including MSIE).
>That seems like a rather cynical attitude. They are asking for
>compatibility with Netscape for the reason that many HTML generators

Yeah, I know why they do it, but so what?  What is the likelihood that
they are Mozilla-compatible and yet don't accept multipart/x-byteranges
instead of multipart/byteranges?  I find it far more likely that a
Mozilla-compatible user agent will suffer the same bugs as Netscape,
if for no other reason than the fact that they code based on
Navigator's input/output rather than reading the specs.  This is obviously
the case for Microsoft, so I see no reason not to assume it for the rest
of the clones.  Besides, it is more efficient to do only one table
lookup instead of two.

>In other words, they are asking for compatibility with Netscape's
>HTML, not Netscape's protocol bugs.

Same thing.

>Or, to put this another way: are you vetoing the patch I introduced
>earlier (which takes out the quotes, adds a check for MSIE, but leaves
>the Netscape-check as is, with Request-Range) assuming I change it to
>"MSIE 3", or are you just complaining about it?

Just seeking improvements (i.e., complaining).  I am not shy about
using a veto when I think it is warranted, but it isn't in this case.
I support the other aspects of the fix.


View raw message