Thanks for the feedback everyone! I'll try to comment on some of the things 
mentioned.

Indeed. I do not know if what you have in mind about coding theory somehow 
> overlaps with combinatorial designs (in particular orthogonal arrays/BIBD 
> :-P) but it would be nice to shake those areas of Sage anyway.
>
What we really feel is lacking right now (with respect to our own research 
areas) is a structured (i.e. object oriented, modular) way to deal with 
families of error-correcting codes, from various angles. Thus, the main 
priority is to introduce such a structure, as well as expected 
functionality for the most familiar families of codes (e.g. those everyone 
meets in a graduate course on coding theory). We are not combinatorists, so 
we know this angle poorly. Please join the discussions about structure etc. 
when we do this on sage-coding-theory 
<https://groups.google.com/forum/#!forum/sage-coding-theory> to make sure 
what we come up with makes sense from that angle!
 

> Make many small tickets. The best way to not have many dependencies is to 
> have your code integrated into the successive beta versions of Sage, this 
> way your 'current code' will never be too far from the develop version.  
>

> Whatever happens, please, don't come some day with 3Mb of changes. Respect 
> our development process: write tickets, get them reviewed. Writing code 
> takes much longer than reviewing it, so if the three of you can work by 
> reviewing the code he writes everything will run smoothly.
>
> Also, each commit message should begin with "trac #number:", otherwise it 
> is a mess to know which commits belong to which ticket when you have many 
> dependencies.
>

Thanks, this is good advice. We are very aware of the threat of patch-bomb 
- this will not happen, I promise. However, in the very beginning (and only 
here) we will play around the OO-structure until we feel sure that it works 
as we would like. We'll discuss details on sage-coding-theory. Once we're 
preliminarily satisfied, we'll brake it into logical pieces and make 
tickets. After that, the new functionality will all be implemented as 
tickets directly.
 

> Once more: please, don't try to shortcut the review process. Play it fair. 
> Don't develop everything in your office only to come back later expecting 
> us to merge everything at once without reviewing it. Please create tickets, 
> please discuss the implementations on the tickets, please review 
> everything. Sage's review process makes things slower, but if you are 3+1 
> to be interested in this project you have the manpower (and the will) to do 
> this properly.
>

We were actually concerned that it would be regarded as too inbred if we 
review David's code? 
 

> I will be glad to help whenever I can. It sure is a a good news.
>
Cool :-D

Working with Vincent on Combinatorial Designs (many tickets between beta 
> releases) taught us that it is better to 'artificially' order all tickets 
> linearly. Each ticket should have a successor and a predecessor, and they 
> should all be linearly ordered. It is tempting to let some branches be 
> independent, but if for some reason you find out later that they still 
> interact somehow it will create a *LOT* of mess if you already have tickets 
> based on those two branches.
>
Hmm, this is a very interesting idea. We'll think about doing something 
like this.

This is great news! Is there any plan to implement semaphore codes in the 
> near future (see the book by Jean Berstel 
> <http://www.goodreads.com/author/show/399140.Jean_Berstel>, Dominique 
> Perrin <http://www.goodreads.com/author/show/926712.Dominique_Perrin>, 
> Christophe 
> Reutenauer 
> <http://www.goodreads.com/author/show/399139.Christophe_Reutenauer> on 
>
"Codes and Automata")? This would be super useful!
>
Sorry, that's not in our road-map currently. The project will be centered 
around error-correcting codes, with a focus on algebraic coding theory. 
We'll also be touching upon the cryptographic applications (e.g. McEliece 
cryptosystem).

Also important: start small, fix a few bugs just to get used to the Sage 
> development model.
>
Yes, this is important. That's on David's todo-list for next week :-) 

Try to make this person attend a Sage Days early on. 
>
Good idea! We'll keep an eye out for one which fits. Also, there is the 
GroupeUtilisateursParis <http://wiki.sagemath.org/GroupeUtilisateursParis> 
which he should join.

The goal should be to have code with *positive_review*, not just to have 
> code. There is no point in continuing to produce more code if there are 10 
> needs_work tickets. 
>
 Totally agree! But there might well be 10 needing_review tickets :-)

I think it would also be a good idea to get in contact with the matroids 
> folk who got a lot of code integrated into Sage some time ago. Or even 
> better with Eric Gourgoulhon whos currently writing a lot of code for his 
> project and getting it integrated into Sage piece by piece. On top of that 
> he's not so far away from Polytechnique...
>
Thanks, we'll look into it, but see my clarifications on the main focus 
above.

Johan

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Reply via email to