Re: filenaming proposal for IAFA description files

Markus Stumpf <stumpf@informatik.tu-muenchen.de> Fri, 16 October 1992 21:27 UTC

Received: from ietf.nri.reston.va.us by IETF.NRI.Reston.VA.US id aa08308; 16 Oct 92 17:27 EDT
Received: from NRI.RESTON.VA.US by IETF.NRI.Reston.VA.US id aa08304; 16 Oct 92 17:27 EDT
Received: from kona.CC.McGill.CA by NRI.Reston.VA.US id aa19848; 16 Oct 92 17:27 EDT
Received: by kona.cc.mcgill.ca (5.65a/IDA-1.4.2b/CC-Guru-2b) id AA04374 on Fri, 16 Oct 92 12:46:44 -0400
Received: from tuminfo2.informatik.tu-muenchen.de by kona.cc.mcgill.ca with SMTP (5.65a/IDA-1.4.2b/CC-Guru-2b) id AA04355 (mail destined for /usr/lib/sendmail -odq -oi -fiafa-request iafa-out) on Fri, 16 Oct 92 12:46:11 -0400
Received: from dsrbg2.informatik.tu-muenchen.de ([131.159.0.110]) by tuminfo2.informatik.tu-muenchen.de with SMTP id <57669>; Fri, 16 Oct 1992 17:45:40 +0100
Received: by dsrbg2.informatik.tu-muenchen.de id <167983>; Fri, 16 Oct 1992 17:45:32 +0100
Subject: Re: filenaming proposal for IAFA description files
Sender: ietf-archive-request@IETF.NRI.Reston.VA.US
From: Markus Stumpf <stumpf@informatik.tu-muenchen.de>
To: Alan Emtage <bajan@bunyip.com>
Date: Fri, 16 Oct 1992 17:45:25 +0100
Cc: iafa@cc.mcgill.ca
In-Reply-To: <9210161555.AA09865@mocha.cc.mcgill.ca> from "Alan Emtage" at Oct 16, 92 04:55:58 pm
X-Mailer: ELM [version 2.4 PL2]
Content-Type: text
Content-Length: 3360
Message-Id: <92Oct16.174532met.167983@dsrbg2.informatik.tu-muenchen.de>

Hello!

|>The name that the IAFA draft suggests for UNIX sites, for this kind of
|>information is IAFA-PACKAGES. We conciously avoided dotted filename
|>extensions because of the problem that you mention. Since the document
|>specifies that you can have more than one record in this file, you don't
|>have to worry about putting more than one package in a directory.... just
|>keep adding the descriptions to the file. 

Yup! I have read this in the IAFA docs, but ...
We have currently indexed 126 packages/documents!
The file containing all the iafa index records 'cat'ted together is
107 KB large!
This 126 packages/documents are really only a small part of our archive
and we are currently not a big archive! (Also we are currently working
on integrating the archives of all universities/departements in Munich
into one big distributed archive).
The problem that arises in keeping more than one index record in one file
is the maintainability!
From the IAFA admin's point of view it's really getting harder to substitute
and/or update existing records for new releases of a document/package.
(Believe me! We are more than one person working on updating the archive!
This means you have to install locking mechanisms to prevent more than one
person/program to update the file and and and).
If you don't keep them in one file you have to have different directories
which doesn't make sense in any case (in most, but not in any!) as two files
cannot have the same name. And having multiple files with the same name
(and this multiple will be around thousands some time) will make it hard to
manage them, even if they are in different directories (on systems supporting
directories)!
Keeping some of the index records together at some location in the (on Unix
systems) directory tree, makes it hard for the user to find them! Especially
to find the file containing the record the user is after (also true for
the admin updating a record).
Another advantage of the - let's call it .iir solution ;-) - is that one
can see at once whether an index record exists for a package/document or not!
Also if a package/document is removed the probability to forget to remove
the index record, too, is smaller if the record is stored along with the
package/document.

From the users view also giving them all the same name, makes it on ane hand
easy for the user to find the index record, but makes it on the other hand
hard for her/him to maintain them, as S/HE has to give them different names
when e.g. retrieving them via ftp or they will be overwritten.

All the thoughts expressed above really did arise from "day-to-day-use".

So, I'd be glad to have both:
the IAFA-PACKAGES file, but also the option to store the index records
in  lots of seperate files with a kinda standardized postfix that makes ist
easy to guess the contents from the name (which would probably be hard
using .TXT).


-----------
Another thing we noticed with the Document abstracts is the lack of a
Language: tag indicating the language the document is written in! I agree
it can in most cases be seen from the Title: or Abstract: tags, but it
would be useful for keyword searches in databases!

Have a nice weekend!
Ciao

	\Maex
-- 
______________________________________________________________________________
 Markus Stumpf                        Markus.Stumpf@Informatik.TU-Muenchen.DE