Re: accept and accept-charset attributes

I no longer think the three accept attributes are particularly difficult 
to implement, at least in Java. Once I sat down and implemented them, 
they were quite straight-forward, and required only minor, non-public 
changes to my code.

I do, however, wonder about the choice of the specific three attributes 
we've been given. accept-language is a clear winner. You might well want 
to choose the French or English or Chinese version of a resource. This 
one makes sense. The use case is obvious and compelling.

accept is more questionable. Still, especially when parse="text", I can 
see asking for the xml or html version of a resource as appropriate if I 
were trying to write a tutorial about one language or the other. It's a 
bit of a stretch but the use-case is there.

accept-charset makes no sense to me whatsoever. Almost always the 
document returned by a server in the requested charset will be a 
straight transcoding of the same document in a different charset. It 
seems very unlikely (and probably a bug) if the document content differs 
from one charset to the next. Furthermore, whatever charset the resource 
is served in, the XML parser will simply transcode into its native 
format. No artifact of the charset should be left after resolution. 
Perhaps the client wishes to ask for only those charsets it knows how to 
process? Perhaps, but this decision is much beter made by the client 
software than the document. For instance, I might wish to code my 
implementation such that it can accept all charsets Java understands. 
Daniel Veillard might wish to code his implementation so that it accepts 
only those charsets libxml understands. I don't see why either of our 
implementations should be controlled by what's in the document. The 
document does not know whether it will be processed by libxml, XOM, 
XInclude.NET, something else, or all of the above. The decision about 
which charsets to accept should be left to the processor, not embedded 
in the document.

Therefore I am making a formal comment requesting that the 
accept-charset attribute be eliminated from XInclude, and that 
processors be allowed to make their own choice of charsets to be 
negotiated with the server.

--
Elliotte Rusty Harold

Received on Thursday, 29 January 2004 16:15:35 UTC