Re: Upgrade Ubuntu Jaunty to Karmic (9.04 to 9.10) breaks self-compiled octave

[Thomas Weber]

On Sun, Nov 01, 2009 at 05:27:22PM +0800, Uwe Dippel wrote:
> Thomas Weber wrote:
>>
>> Off the top of my head (and maybe mixing some version numbers in the
>> 2.0/2.1 case).
>>
>> When 2.0 was stable, it was so for a really long time. Most users
>> actually used the 2.1 version. So, both were introduced into Debian
>> (I think this was actually done by Dirk Eddelbuettel, but it happened  
>> before I got involved in Debian, so maybe it was someone else).
>>
>> 2.1 vs. 2.9/3.0: There were some big changes, breaking other people's
>> software. I don't remember what exactly, but I submitted at least one
>> patch to another maintainer to get his package working on the new 3.0
>> stuff.
>>
>> Ultimately, we are talking about Octave as a library and an interpreter,
>> not as a standalone software.
>> If a new release isn't 100% backwards compatible, you somehow need to
>> give people a way to migrate their scripts to the new version. That
>> means they need the old release in ordr to work and the new release for
>> porting their stuff.
>>   
>
> Dear Thomas,
>
> appreciate your explanation very much. I thought so, since Eidors was  
> broken with 3.0. Unfortunately, 2.X was not available under Jaunty, and  
> some list members suggested to compile 3.2.2 myself. (Which was quite a  
> hassle, but finally done with some very helpful people on this list.)  
> The whole thing broke again, with the update of Jaunty to Karmic.

I'm sorry to say that, but you are using the wrong distribution. Get a
distribution with a *real* maintainer caring for Octave. Ubuntu's "let's
recompile whatever happens to be in Debian unstable/testing at an
arbitrary point of time" attitude just doesn't cut it. That goes for
Octave's dependencies as well.

If you don't want Debian, Fedora seems to be a sensible alternative.

> Maybe I should say a few words about my background and motivation: being  
> a FOSS-person for > 10 years, I was looking to replace expensive/pirated  
> Matlab stuff in my workplace (students can easily buy Matlab for around  
> US$1 in various markets). We were doing EIT-research, using Eidors. Our  
> data, alas, cannot be shown on octave, subsequent to the transition  
> 2.X->3.X of octave, because it broke Eidors compatibility; so we are  
> stuck with Matlab. 

> Add together with my hours (now rather days) just for installing, I'm  
> not really impressed. Though I tried very hard to be impressed. For next  
> Friday, I am invited to show the potential replacement of proprietary  
> research tools (SPSS, Matlab) with FOSS tools in the MyGOSSCon 2009  
> (Malaysian Government Open Source Conference). What should/could I say  
> without lying? Oh, yes, qtocatve is incompatible with 3.2.2.

Hm, are you sure about the incompatibility? I know for a fact that 3.2.3
works:

https://buildd.debian.org/pkg.cgi?pkg=qtoctave

> Also in this respect, your sentence
>
> Ultimately, we are talking about Octave as a library and an interpreter,
> not as a standalone software.
>
> confuses me. Does that mean octave does not want to be a competitor to  
> Matlab?

I don't know what Octave wants to be. I don't care about Matlab at all.
I'm no longer at the university and buying Matlab is out of the
question.

My sentence was meant in the idea that Octave is not a stand-alone
project like OpenOffice.org. Ie, I need to care for people having
written code specific to older versions of Octave by giving them a
migration path.

> I really don't know what to say at MyGOSSCon. As of now, I can only  
> summarise, that I couldn't find fault with R as a full replacement of  
> the functionality of SPSS (nevermind GUI), function; while octave -  
> sorry, only to my own personal experiences in the last few months, and  
> not wanting to hurt anyone - looks more like alpha. At least, seen from  
> the perspective of someone who intends to replace it for Matlab in  
> teaching and research.

<shrug> I've used Octave for several years at the university without
problems. If it doesn't meet your needs, you seem to have different
needs than I have.

> (And, from the point of view of a FOSS-developer, a software that seems  
> to have seen some wrong decisions. But that's not what I feel like  
> discussing here. Terms like backward compatibility and Matlab  
> compatibility would need addressing before I could full-heartedly  
> recommend its use.)

Then continue to use Matlab. I don't see a problem with that at all. If
money is a problem, <shrug>. You can't eat your cake and have it, too.

        Thomas