[daip] AIPS user data areas

[computer support]

>Date: Mon, 13 Jan 2003 11:01:33 -0700
>From: Eric Greisen <egreisen at cv3.cv.nrao.edu>
>To: computer support <help at astro.ucla.edu>
>Cc: daip at primate.aoc.NRAO.EDU
>
>computer support writes:
>
> > I'm surprised, because I thought that at least one disk had to be 
> > required ("+" entry), based on instructions read long ago.  Also, 
> > we found problems if that disk was not writable by any AIPS user 
> > ID ("no disk available" or something like that)...
>
>     In the old days of a few public "big" machines and no private
>ones it would help to have a required disk as disk 1, where the users'
>SAVE/GET and message files would reside.  Normally now each user has
>his or her own machine.  If that user stays only on his/her machine
>then things are very much faster and no one steps on anyone else and
>disk numbers do not change unless the disk has crashed.  Note that it
>takes about 1 second wall time to write one message to the aips
>message file over NFS.  1000s of messages may be written in that time
>if the disk is local.
> > 
> > Her problem was that, if a disk ahead of hers in our list was 
> > unavailable, the number of her disk (and therefore her catalog 
> > file names or whatever) would change and she wouldn't find.  
> > It seems that making no disks required will make this worse, 
> > as the numbering will now vary even more, depending on what 
> > hosts are given to the "da=" command line option.  
> > 
> > But maybe I'm missing something.  We'll try it that way...
> > Anyway, thanks for your prompt reply on Friday.
> > 
>
>Users rarely invoke the da= option and those disks always follow the
>local disks and so do not affect the numbering for the more important
>local disks.  When a random collection of machines are all included
>then the user will get confused because they have no idea what the
>full set of machines is doing.  They should know what 1 or 2 machines
>are doing if they choose to include them explicitly via da=.  As I
>say, most users here do not even know that that option exists.
>
>Eric Greisen

The problem here is that just about everyone has their data spread
over disks on their personal workstation and the (remote) PUBLIC 
disks (for historical reasons).  Furthermore, students usually run
AIPS from their own workstations, which do not have any local disks.
And, even users who have disks on their workstation may run AIPS
from other systems sometimes.  So the disk number assignments will 
be completely different for everyone (and even for a given user, 
depending on where they're running), and they will have to specify 
a varying list of several systems when starting AIPS.

We'll see how that works out...

Philippe