[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [syndication] Re: My take on shared feed lists
> >it up as text/plain or text/xml.
>
> Um, both of those and especially the second are valid aren't they? The
> problem is that the types we need to define are a layer above this. eg
> type(MIME)=text/xml and then type(content)=RSS 2.0
Well, it'd be perfectly valid to call it binary/octet too as it's really
nothing more than just bytes. But that would be silly. It's not nearly as
silly using text/plain but it's certainly not as ideal as one might prefer.
Here's the thing, in order for a site to *correctly* serve out data following
a fully compliant HTTP it will need to have the server be involved in altering
the header during connection negotiation and/or delivery. This is entirely
unrealistic for the staggering majority of current sites with feeds. This isn't
some grave disaster but it would help if they used a more sensible type.
So making use of a extensible type would accomplish two goals, one is to let the
application know what form of data is available and two, prepare for consuming
it. As has been pointed out, this type is no 'guarantee' but it's certainly
better that just winging it and trying to 'guess' at how to parse it.
Think of it this way, a resource limited device (ram, network, cpu, pick any) on
seeing the
head section info could decide to make a HEAD request soley on the type it knows
it can support. Should that HEAD request return a size that it knows it can
manage
then it would request the data. This as opposed to making a blind request and
getting too much data in a format it can't use.
> Looking at the html4 spec, there's a mechanism to define new
> rel=link-types. So we could say rel="meta" or rel="metaindex" as long as
> that is defined in a <head profile="URI">. The problem here is
> mime-types. We've been playing fast and loose with mime types for years
> but it still hasn't got us to application/xml+rss_2.0
Fast, loose, fractious and lazy perhaps sums it up better. MIME types, we all
love 'em, we all fail to do the diligence on them. That said, damn the
torpedoes, full steam ahead!
Thus another reason to /allow/ for using a URI in this attribute. After all, we
know what we're looking to find and using a URI gives us a chance to document
what format it's using. Putting up a web page at a URI works wonders for
self-documenting the format being used.
-Bill Kearney