Re: SVG12: image/svg+xml gzip requirements

Boris Zbarsky wrote:

> Thomas DeWeese wrote:

>>    I don't think there is generally any other way to detect
>> inconsistent meta-data is there?
> 
> Sure there is, when there are actually two sets of metadata.  You just 
> compare them.

   Well, in this case the other set of metadata is inside the file.

>> Heck even within XML the detect of UTF-8/16 relies on this basic 
>> technique.
> 
> This is only needed in XML because the metadata cannot be read without 
> this indication, basically.  And that's a consequence of not having a 
> place to store metadata outside of the main data in typical filesystems...

    Yes, well "self describing" content must bootstrap it's self.
To be honest I'm not that interested in getting to an argument about
is this good or bad or where the line should be drawn, but there is
lots of "self describing" content that is sniffed in this manner
all the time and it is in practice extremely reliable.

>>> This all sounds to me like an attempt to legislate existing broken UA 
>>> and content provider behavior into technically being correct.... 
>>
>>    Who said any of this was a matter of "correct".
> 
> We're talking about what sort of behavior is considered conformant 
> (equivalent to being correct, in my mind, when talking about 
> specifications) to the SVG specification.

    Actually the TAG finding (which _is_ what I was responding to)
clearly views this as a UA working around a mistake.  Hence the
need to notify the User.  The question is if this should be allowed.
I would say that if the UA notifies the User it is 'conformant'.
I guess you would say it's not (I _might_ even be willing to go
as far as saying the UA is not being conformant but it's still
"correct" from a standards perspective because it's telling the user
it's not being conformant).  I really only got involved because
of Bjoern implied that the TAG finding did not support UA's
detecting simple errors and working around them, and it clearly
does (with some provisions).

> Where did I ever mention TAG?  Please cite such claims.

   My apologies, that was Bjoern.

Received on Wednesday, 8 December 2004 20:53:02 UTC