home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!spool.mu.edu!umn.edu!news2.cis.umn.edu!gopher-news-daemon@boombox.micro.umn.edu
- Date: Thu, 19 Nov 92 18:48:09 PST
- From: barries@nye.nscee.edu (Fred Barrie)
- Message-ID: <9211200248.AA24641@nscee.edu>
- Original-To: gopher-news@boombox.micro.umn.edu
- Subject: Status report on Veronica development.
- Newsgroups: comp.infosystems.gopher
- Distribution: comp
- Sender: news@news2.cis.umn.edu
- Approved: comp.infosystems.gopher@news.cis.umn.edu
- Lines: 87
-
-
- Status report on Veronica development. 11-19-92
-
- 1. Imminenent enhancements to Veronica:
-
- A. Veronica should return one and only one reference to each
- distinct item on the Net. We are experimenting with several
- approaches and will implement a trial of at least one method,
- in the next few days.
-
- The Veronica database will be rebuilt to give faster accesses
- with the redundancy-removing algorithm.
-
- 2. Preliminary thoughts on proliferation of Veronica servers.
-
- A. Each organization possessing several gopher servers should
- be encouraged to run a site-wide Veronica service for those
- gopher servers.
-
- B. There should be "several" Veronica servers which index all
- of gopherspace, possibly excepting the most transient data
- ( see C below ). Thusfar, our little NeXT Veronica at
- Nevada has not crashed under the load ( typical "load"
- between 3 and 8 ). But it's response is far from fast enough.
- These large Veronicas need faster hardware.
-
- C. Specialized Veronica servers for transient data. USENET
- offerings in particular need not be indexed and reindexed
- by all the Veronica servers. We have found it difficult to
- automatically filter USENET news directories and to
- eliminate redundant presentation of USENET items. Total
- Veronica loads could be reduced by implementing specialized
- Veronica servers for such transient data.
-
- D. Some Veronica servers could specialize by topics. This could
- be implemented easily if these servers got the relevant
- data by ftp from the 'large' Veronicas of B. above.
- We think it would be much more effective to acquire the data
- from the large servers, rather than propagating 100s of
- gopherwalks all over the net.
-
- In general we are wary of this proposal. The utility of
- Veronica derives from its ability to index "everything".
- Topic subdivision is only as good as the keyword scheme.
- We think those schemes are never very good. It is much
- better to have a fast index of "everything" than to have
- a very fast index that has been subjected to someone else's
- preconceived categories.
-
- Several early commentators have suggested that topical
- division are important to Veronica's utility. Most of
- these comments objected to a volume of seemingly irrelevant
- items returned by the searches. We think this volume will
- be much reduced when we elimate duplicate items. Further,
- we think that much of the annoyance was caused by WAITING
- for duplicate and seemingly-irrelevant items. We think
- that faster searches without redundant items can satisfy
- most users while still indexing "everything".
-
- We recognize that we may repent this opinion if (when) the
- volume of gopherspace explodes.
-
- Meanwhile, we suggest that the following additions to
- the gopher+ spec could make Veronica more efficient.
-
-
- 3. Suggested additions to gopher+ protocol.
- Add some information to the new gopher-item fields.
-
- A. An "indexing" field ( boolean ) could be added, perhaps
- to the ADMIN block. A Veronica could ignore items so marked,
- such as News directories.
-
- B. A "subject=" field could be added. This would rationalize
- the topic-oriented searches discussed in 2-D above.
- "Subject" must take multiple values. We recognize that this
- is a watered-down ABSTRACT.
-
- Items lacking a "subject" field would inherit "subject"
- from their parent.
-
- C. A boolean "local" field. This would help prune the
- duplicate references.
-
-
- Fred Barrie, Steve Foster
- University of Nevada
-