Great Circle Associates Majordomo-Workers
(February 1997)
 

Indexed By Date: [Previous] [Next] Indexed By Thread: [Previous] [Next]

Subject: Re: Why must I create twice the number of directories?
From: Arnaud Taddei <taddei @ mail . cern . ch>
Date: Mon, 17 Feb 1997 12:58:34 +0100 (MET)
To: Jason L Tibbitts III <tibbs @ hpc . uh . edu>
Cc: majordomo-workers @ GreatCircle . COM
In-reply-to: <ufak9obm1a8.fsf@sina.hpc.uh.edu>



On 14 Feb 1997, Jason L Tibbitts III wrote:

> >>>>> "CMG" == Carlos M Gutierrez <carlos@gutierrez.com> writes:
> 
> CMG> How about using DB-style databases, but the additional information
> CMG> would be stored separated by '|'s
> 
> Well, that's a possibility, but it's a pain to then have to deal with
> quoting.
> 
> CMG> $user{'luser1@luser.com'} = "fullname|digests|var3|var4|varN|";
> 
> And if any of those fields contain a '|'?  Or any other character you might
> choose?  Tabs are probably a good choice because you can safely smash them
> to spaces in all cases we can control, and they aren't legal in all cases
> we can't control (email addresses + comments).
> 
> CMG> or is this similar to "MLDBM"? (not familiar with it)
> 
> MLDBM lets you store complex data structures (lists of lists of hashes of
> hashes of lists of scalars, say) as the values of a tie'd hash.  It's just
> a cute hack using Data::Dumper behind the scenes, which is why it would be
> great for some things but might end up being pretty bad for what we need.
> Someone needs to do some testing.
> 
> CMG> One could have secondary indexes to speed some lookups, but
> CMG> synchronization and locking between 2+ dbm files becomes a tricky.
> 
> Especially if you assume that the machine will crash in the middle.
> 
> The real problem with databases is that people can't poke about in them,
> and rest assured that people want to do so.  I suppose it's easier to start
> with a flat file format having the data fields and then decide what to do
> about a database later.  Right now I'm trying to operate at an even more
> basic level of duplicating the 1.9x functionality.
Jason, you have already my input and our address.tab file looks like very
much your flat file. Having the separator as "|" instead of my "=" is fine
for me. Now, it is clear as well that when your directory is big enough,
having the data in a flat file is not enough, becasue of speed concerns.
This is why I hope that we can use an ldap service with all the data stuck
into memory. This would make all lookups much faster and it integrates
very well with other mail services components like the mail agent
(Netscape, etc.) or sendmail (see READMEs of sendmail 8.8.4 +).

Having said that, I accept very well that you start with a simple flat
file.

-----------------------------------------------------------
Arnaud Taddei		tel:  +41 22 767 9349
IT Division  513 1-019	fax:  +41 22 767 7155
CERN			mail: Arnaud.Taddei@cern.ch
CH-1211 Geneve 23	URL:  http://wwwcn.cern.ch/~taddei
-----------------------------------------------------------



References:
Indexed By Date Previous: making web interactive with majordomo
From: Patrick Fitzgerald <fitz@iquest.com>
Next: date_info
From: Brock Rozen <brozen@webdreams.com>
Indexed By Thread Previous: Re: Why must I create twice the number of directories?
From: Jason L Tibbitts III <tibbs@hpc.uh.edu>
Next: Digests and MIME
From: "James B. Byrne" <byrnejb@harte-lyne.ca>

Google
 
Search Internet Search www.greatcircle.com