I have an opinion about the technical solutions to this problem, but
I need to say a couple of things up front, since this has the
potential to be such a political issue:
First, a disclaimer: All of what I'm about to say is my personal
opinion as a Mac user and software developer who's done a lot of work
with XML and syndication formats, and not necessarily the opinion of
my employer. While UserLand makes software that produces RSS 2.0 and
consumes Atom and RSS of all flavors, what I want to advocate is that
we try to take the politics out of the software as completely as
possible, and focus on giving the users and content providers the
best experience and the most choice.
I strongly urge everyone on this list to refrain as much as possible
from bringing the RSS vs Atom politics (and its inevitable flame-
fest) into this yet again. Personally, I've really had it with
watching so many smart people bicker with each other about it (and
sometimes much worse). I know for certain that many other developers
feel as I do in this regard.
Having said that, let's try to get some facts about early
autodiscovery development out in the open.
I believe that the original work on the autodiscovery feature for
syndication feeds predates Atom, and in fact specified that feeds
were RSS. This should be relatively easy to corroborate with a little
reading via Google and archive.org.
As far as I can tell, the original suggestion for this use of the
HTML link tag was made by Mark Pilgrim in May of 2002. In early June,
2002, early adopters of the feature agreed to a change to the values
of the title and type attributes to make the links more specific,
however the feature has changed little, if any, since then.
I've seen no mention of Atom with regard to this early work on
autodiscovery from around this time. Here are some relevant links:
Later on -- and I don't recall exactly when it started -- a major
sh*tstorm erupted between some of the people involved in RSS and some
of the people involved in originally designing the autodiscovery
feature, who were now in the Atom camp. There's no need for us to go
into detail about the rift here, nor will I take sides, but clearly
we're now at a point where it's become important to address the issue
in our software.
Some time after this sh*tstorm, the autodiscovery RFCs (specs?) from
Aug 2004 and May 2005 showed up, both of which appear now as part of
the Atom IETF draft which came long after the initial proposals:
By Aug 2004, the link tag had been in widespread use for RSS feeds
for *over two years*, and these days, as we're well aware, it's used
for both RSS and Atom feeds. In my opinion, both uses are equally
legitimate. It's certainly true that both conform to the HTML spec,
and at least in spirit, both uses conform to the original intent of
the (now politicized) autodiscovery proposals by Mark Pilgrim and
others.
I'll posit that there are probably only two acceptable solutions to
the problem of which link to pick up with autodiscovery:
* Option 1: Give the site author the control. As far as I know, most
sites which have both feeds present a link tag for each. Certainly
many (most?) of these sites will be authored in order to express the
content provider's preference for one format over the other, by
presenting either the RSS or the Atom link first. Have Safari respect
the link order, and use the first feed that it finds.
* Option 2: Give the user a preference setting: keep the decision in
the hands of content consumers. This preference can live in Safari,
in System Preferences, both, or wherever is appropriate. There is
prior art for this in preferences for which application to use for
HTTP, FTP, Mail, etc.
The preference *should* be able to be changed (with user
confirmation) by an external feed reader application. Given a
preference setting, the default feed reader -- Shrook for example --
could check its value, and warn the user that they should (allow
Shrook to) change it to prefer RSS instead of Atom feeds, whenever
both feeds exist. The user gets the content they want, in a format
that Shrook can read, without any substantial change in Safari's UI,
and without worrying about what side of the syndication format wars
the content provider happens to be on.
It's worth pointing out though, that if Safari were to go with the
user preference solution, then there is a looming issue of what the
default value is for the preference itself. I would argue that since
iTunes is already leaning towards RSS 2.0 with its Podcasting
support, it wouldn't be nonsensical for Apple to default to the RSS
feed, but this a decision with political ramifications which are
similar to the current behavior of preferring the Atom feed.
In the end, any solution which doesn't allow either the users or
content providers (or both) to choose a preferred format is
unacceptable in my opinion. Anything else will likely fan the flames,
or at least appear clueless on Apple's part.
It's quite possible that in the end, the best solution will be a
combination of the above, or something else that's yet to be
described -- However I'll stick to my core opinion that the choice
*must* be left in the hands of users and/or content providers, or
else more flaming and wringing of hands await.
-Jake Savin
ps. Sorry if anyone received this message twice. This is a re-send,
since the one I sent yesterday didn't appear to go through.
Jake Savin - Lead Developer
UserLand Software
www.userland.com
