Re: atompub-multipart-01: browser implementation

[Aristotle Pagaltzis <pagaltzis@xxxxxx>]

* Bill de hOra <bill@xxxxxxxxxx> [2008-06-20 13:05]:
> Aristotle Pagaltzis wrote:
>> In this case, surprisingly, I think the boolean attribute is
>> accidentally helpful: with a non-boolean attribute, care would
>> have to be taken to allow multiple values in the future when
>> more than one way of multipart processing is specified and
>> collections can support several of them. With boolean
>> attributes, you just make up more of them.
>
> <atom:link related="true"   ... />
>
> Don't think anyone here would buy that.

You snipped the paragraph where I said this is not a laudable
approach.

Anyway my overall point is somewhat bigger and I’m failing to
convey it. Let me try again.

The boolean markup should be excised, no argument.

However, I think it won’t matter that much, bottom line. We have
far too little experience with packaging support in collections
to know how to how to advertise it in the general case, and even
less information about how to make multiple advertising clauses
play together plausibly. Making the markup in this spec more
palatable won’t make it any more likely that whatever we come up
with will prove adequate for future concerns.

Out of our position of ignorance we should do the simplest thing
that can possibly work. (And boolean markup does *not* work.)

To me that also means doing it in as straightforward a manner as
possible over HTTP (very much the spirit of RFC 5023); and that
means using `multipart/related` in preference to anything that a
contemporary browser can natively send.

Regards,
-- 
Aristotle Pagaltzis // <http://plasmasturm.org/>