Re: rel="discuss"

[Aristotle Pagaltzis <pagaltzis@xxxxxx>]

* Sylvain Hellegouarch <sh@xxxxxxxxxx> [2008-05-10 17:50]:
>> I am asking what would be gained by having
>>
>>     <link rel="conversation" href="xmpp://foo"
>>         label="Chat about foo" />
>>
>> that it could not just as readily infer from
>>
>>     <link rel="related" href="xmpp://foo"
>>         label="Chat about foo" />
>>
>> seeing as the critical bit of information for the machine
>> seems to be conveyed by the `xmpp:` part in both cases and the
>> critical bit of information for the user would be conveyed by
>> the label in both cases.
>
> It seems to me that having to parse the URI to decide whether
> or not such a link is usable by the program reading it is
> fairly poor design decision.
>
> From a machine POV, how would you differentiate those:
>
> <link rel="related" href="xmpp://foo"
>         label="Chat about foo" />
>
> <link rel="related" href="xmpp://bar"
>         label="Monitor bar" />

Do you actually need to differentiate them? If you’re merely
going turn these links into buttons for the user to click, with
the `label` as the label, then I don’t see any need for the
machine to be able to tell them apart.

> The second might not even interest the user per-se, it might
> just be targetted at auto discovery or what have you.

Yes, well, then the second link (but not the first, funnily
enough) needs a more specific link relation so to facilitate
autodiscovery by automatic agents because that one is intended to
be processed automatically. That’s exactly what I’ve been saying.

> You could make the URI more explicit probably but why refusing
> the use of the rel attribute in such case anyway?
>
> I'm trying hard to understand your reluctance by gauging how
> that would hurt any out of band context but I'm unsure it's the
> right battle.

If you use `related`, there are quite a few agents that will
consider this a link that might be interesting to a human, and if
you use standard URI schemes, many of them will be able to direct
clicks to the right application. Eg. if you put this in a regular
HTTP-served newsfeed:

    <link rel="related" href="irc://example.org/project-foo"
        label="Tell us your opinion on #project-foo!" />

then it’s quite likely that web-based aggregators will turn this
into

    <a href="irc://example.org/project-foo">Tell us your opinion
    on #project-foo!</a>

and place it somewhere in their item chrome, and whaddayaknow,
clicking such a link in a browser will actually launch an IRC
client and take the user to the channel.

So what Peter thinks he needs a new relation for already works!
(If I understand his use case correctly, anyway. So far he has
not been forthcoming with qualifications.)

In contrast, if you use a non-standard relation, this will *not*
because clients will generally ignore links with relations they
do not understand. So if it were

    <link rel="conversation" href="irc://example.org/project-foo"
        label="Tell us your opinion on #project-foo!" />

then only few Atom processors will actually do anything useful
with that information.

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