On Wed, 5 Dec 2007 21:11:21 -0500, Farrukh Najmi wrote
> ATOM 1.0 is an IETF standard and has mass market adoption because of
> its simplicity and extensibility.
Remember, ISO 23950 is an ISO standard:-) :-) :-)
Mass market adoption? You want that.. RSS 2.x has a larger market share
in blogs, cms and news feeds by a VERY large factor. So?
> These minor issues are being worked on in the next version of ATOM
> where we could provide useful and constructive input based upon
> practical experience. However, none of this justifies reinventing
> the wheel with our own format.
Its missing A LOT and if we then just pile it into Atom via namespaces I
don't see the point (we could do the same with RSS)... worse still.. what
it has.. needs to, I think, be twisted to fit like Cinderella's shoe...
> I am sure that we can find quite a handful of warts in the current
> SRU response format (e.g. lack of even an option if author etc.)
> Frankly I see missing these important attributes (particularly id)
> in an information management context to be sacrilege. Having these
> as required may be over kill but is is certainly better then underkill.
Lack of persistent Id is not a sacrilege but I'd suggest demanding a
mandatory one that is "unique" and has a semantics...
Lets look a some REAL (existing and in the works applications)-- no
theory, not vapour but real.. (Don't think about the systems 10 years ago
but you should start with SRU to think about what many systems might look
like in 5 or 10 years)..
The problems get even more twisted, I think, when we start to talk about
In a news system its relatively easy.. just return what the article claims.
For a book or a journal in a library system it too is easy.. But in the
What about the general case (which I'd like us to pursue) where we start to
have volatile search directed units of retrieval: where an item (unit of
retrieval) is a "blob" of information deemed by a system as an appropriate
unit of retrieval for the information request.
Its about finding "relevant information" in a sea of information. We need
to-- and increasingly we are-- transcend the notion of object and think
in the kind of informal terms humans use when they talk about information.
If I go to a lawyer and ask about export taxation I won't be referred to the
whole code but to a paragraph in one law, perhaps an EU memorandum and maybe
some sections in various discussions on case law.
In a "static" work such as Shakespeare's plays (where we already apply such
techniques) a unit can be a line, speech, or a group of speeches, an act
or perhaps the whole play or maybe even a collection of plays (viewing plays
in context of a collection).
In a collection of legal documents it can be a paragraph, a section perhaps
a code or a collection, perhaps just a reference.
For the item in a set of plays we can talk about the 2nd act of so and so..
In a collection of legal documents we can address down to the level of
paragraph in a consistent manner.
But what about volatile collections of information where the information is
Sure we have Ids but they are hardly longer term persistent.. not really
the stuff for real URIs.. and from a functional perspective why should it?
Fake URIs? Sure we can cook them but why? Fine when for some reason we
want to make an atom page.. or a RSS page.. we can be "creative" and invent
stuff to seem plausible.. And who cares if I lie and they're not persistent!
Heck, a large number of the news feeds we process are not 100% kosher and
worse still are often non compliant with the standards they think they
But as the real result format should we really leave people to invent their
own "meaningless" ids? Or just set new semantics for persistent to mean
"as long as I feel like it" (like 9ms)..... Unique as in.. throw in a
What is to gain by all this? Do we really gain wider acceptance by trying
to pack this all down the throat of Atom? Like foie grois I think we
spoil a good bird but unlike it we don't get tasty liver! Nothing wrong with
producing an Atom or RSS response--- we do so as well-- but as the basic
If I'm looking just for something "mainstream" (and something that fits
common paradigms of how the mainstream think today it wants to search) then
why should I bother with SRU and not just opt for OpenSearch?
Edward C. Zimmermann, Basis Systeme netzwerk, Munich
Office Leo (R&D):
Leopoldstrasse 53-55, D-80802 Munich,
Federal Republic of Germany