RE: Issue: JW7 (redirects)

At 03:31 PM 6/7/99 -0700, Yaron Goland wrote:
>In many cases a search arbiter may not [kn]ow the full response to a search
>query but may know where to redirect the searcher to find them. So you could
>make a single query to the search arbiter, get back some responses and also
>get back an instruction that you should repeat the search at 4 other places
>in order to get the full response. 

For sure I agree that an arbiter may not have full knowledge.  But...

1) The "search redirection" (2.6 in DASL) doesn't allow an arbiter to both
return partial results (what it does know about) and also a pointer to
additional sources.  Redirection in DASL is indicated by a 400 Bad Request,
with an XML body that lists one or more alternative sites.  So either the
request succeeds (and all you get is whatever knowledge that arbiter had)
or it fails.  You can't get both.

2) Also, I don't quite get how an arbiter could *know* that it has only
partial  knowledge.  Either it recognizes the scope (which is a URI) or it
does not.  (I suppose you could have very smart arbiters that recognize
that they don't index certain properties, but know of another than does,
but that seems implausible.)

DASL's search redirection is supposed to help with cases where the search
arbiter knows which scopes its supports, and also knows the locations of
other arbiters for certain other scopes.  So e.g. if I try to search at
ietf.org with a scope of w3c.org, the nice arbiter at ietf could say "no,
go away, but try w3c.org".  However...

3) I claim that it will be very rare indeed that arbiters operated by
independent organizations will be set up to cooperate in this way.  I mean,
if I search at shell.com  with a scope of chevron.com, do you think
shell.com is going to help me?

4) Even in those cases where servers are able and willing to be nice, the
*only* advantage that DASL's search redirection provides over the already
existing 301 reply is that it allows for multiple URIs in the response.
This additional functionality is not worth the trouble.

Therefore I propose we drop scopeerror and redirectarbiter.

DASL's current design won't do what you asked for, and the existing
redirect facility meets (I claim) most of the likely and plausible need.

Received on Wednesday, 9 June 1999 16:00:08 UTC