Re: [tablix-list] gtablix without clusterconf error

From: R. Quenett <>
Date: Wed Feb 23 2005 - 15:20:26 CET


from: Tomaz Solc

" 0.2 branch of Tablix will be much more flexible once it is finished.

I share Boštjan's anticipation.

" There is an example version 0.2 configuration file on the Wiki. Among
" other things, it will allow you to name each day.

Great. I imagine this option might go under the <resourcetype
type="time"> tag, something like
 <day labels="label1","label2",...,"label(nrDays)" /> ? Btw, I
really like the cycle/slots terminology.
" On the other hand I don't think a central configuration file for Tablix
" would be a good idea. What settings exactly would be stored in there? I
" can't think of many settings that do not change between different
" timetabling problems.

I must confess to not having given it a great deal of deep thought
as to exactly what would be stored in there. You have to know that
I am really neither very skilled with computers nor knowledgeable
about linux. However, many linux programs seem to go in that
direction (it's sort of a quasi-standard?) and I've become used to
looking there when I set out on one of my tinkering episodes. Some
candidates might be anything that could be specified on the command
line (which would override if present) or some of the things that
Boštjan mentioned - string values etc. One thing I have noticed
about myself and some of the very small amount of very bad code that
I have written (and some of it was in BASIC, so I'm ruined
forever;*) was that I had, and still do have, a tendency to hard
code a lot of things that were a pita to change later, as I often
wound up wanting to do. So I've taken to .cfg files and the like
when I can figure out how to implement them.
" PVM_ROOT environment is required by PVM, not Tablix. It is also only
" required when PVM is installed by hand. Binary packages usually don't
" need it (for example Debian PVM packages). I don't think code for
" detecting this should be in Tablix.

I generally like the linux approach of one program doing one small
thing, and only that, very well. So I sympathize with what you say
and it's your decision, of course. However, this approach also has
been at the root of one problem (my learning enough of the overall
massive amount there is to learn about how various programs
communicate and interoperate so I can get started actually using
stuff) that I've had gaining some slight familiarity with linux and
I think it's possible to take the kiss principle too far. By way of
analogy, consider a railway which took an extreme minimalist
approach too far. Embarking passengers is a very different
operation than disembarking passengers so it's very clear that
passengers must never be allowed to get off the train at a station
where passengers are ever allowed to get on. That keeps things
clear, separate and simple (or maybe sometimes not so simple:).

In the same sort of way, for example only, you are completely right
about PVM_ROOT. It's required by PVM, not Tablix. However, I
couldn't get tablix to run until I set it and the cryptic (to me in
my ignorance) error messages weren't much help. Maybe that's just
the way I had/have things fouled up here but perhaps also it's
required for communication between tablix as normally configured and
pvm - in which case it becomes a little less clear as to which
program it is that requires it. In any case, it took me a few hours
and a few gallons of coffee to figure out that's what I had to do.
I learned a couple of things, too, and that's not a bad thing. But
time is a limited resource.

Received on Wed Feb 23 15:31:14 2005

This archive was generated by hypermail 2.1.8 : Tue Aug 16 2005 - 20:43:06 CEST