Collecting ideas for posets

[Jori Mäntysalo]

How should ideas about posets and lattices be collected? (In general:
about some topic X on Sage.)

For example there should be a function to check if a poset is
series-parallel composable, a function to see if a lattice is vertically
indecomposable, an iterator over distributive lattices... And those can be
added to trac or discussed in this list.

But I think that we should collect something to wiki or somebody (me?)
should make a PDF. Then we would have bigger picture of what we want.

--
Jori Mäntysalo

[Nathann Cohen]

How should ideas about posets and lattices be collected? (In general:
about some topic X on Sage.)

My opinion is that the only thing you can trust is a ticket in needs_review. Everything else risks being forgotten, on some 5yo trac ticket or an unmaintained wiki page.

https://groups.google.com/forum/#!topic/sage-devel/xMbCDY6VGuU

Nathann 

[Jori Mäntysalo]

On Mon, 22 Jun 2015, Nathann Cohen wrote:

>> How should ideas about posets and lattices be collected? (In
>> general: about some topic X on Sage.)

> My opinion is that the only thing you can trust is a ticket in
> needs_review. Everything else risks being forgotten, on some 5yo trac
> ticket or an unmaintained wiki page.

True, but what about general "where to go" - discussion above the level of
one function? It trac even good for something like this fresh ticket:
http://trac.sagemath.org/ticket/18757 ? Real question is not how to do it,
but what to do.

--
Jori Mäntysalo

[Nathann Cohen]

> True, but what about general "where to go" - discussion above the level of
> one function?

I'd say that we usually ask them here.

> It trac even good for something like this fresh ticket:
> http://trac.sagemath.org/ticket/18757 ? Real question is not how to do it,
> but what to do.

Are you fearing that you might implement the code that you propose on
the ticket (have them contains "complex" objects as points) only to
find reviewers that do not like the idea?

It looks rather harmless. And you already have a defense ready (if
needed) with your 'elements' keyword.

Nathann

[Jori Mäntysalo]

On Mon, 22 Jun 2015, Nathann Cohen wrote:

>> True, but what about general "where to go" - discussion above the level of
>> one function?

> I'd say that we usually ask them here.

OK. I'll write some notes to myself and ask about them one-by-one.

>> It trac even good for something like this fresh ticket:
>> http://trac.sagemath.org/ticket/18757 ? Real question is not how to do
>> it, but what to do.

> Are you fearing that you might implement the code that you propose on
> the ticket (have them contains "complex" objects as points) only to find
> reviewers that do not like the idea?

It's no big deal in a function or two like here, for bigger changes it
might be. But always I don't have clear opinion about what to do, just
that something should be done.

--
Jori Mäntysalo

[John Cremona]

Sometimes this sort of discussion happens on one of the more
specialsed mailing lists, such as sage-nt, sage-algebra -- or
sage-combinat? Perhaps the people active on that list can say whether
this discussion might be better there than on sage-devel.

John

[Nathann Cohen]

> Sometimes this sort of discussion happens on one of the more
> specialsed mailing lists, such as sage-nt, sage-algebra -- or
> sage-combinat? Perhaps the people active on that list can say whether
> this discussion might be better there than on sage-devel.

I was about to say that "such forums tend to restrict to a subset of
developers some conversations that have an impact for everybody".

I just loaded sage-combinat's forum, to find out that there is a
question there about graphs, which of course are not part of combinat
>_<

Nathann

[Simon King]

Hi Nathann,

I think there is no "sage-graphs" forum. Neither is there
"sage-topology". And if you don't consider graphs as a graphs or
topological spaces, then they appear to be combinatorial objects...

Cheers,
Simon

[Nathann Cohen]

> I think there is no "sage-graphs" forum.

There is none indeed. Graphs are just part of Sage's code.

> Neither is there
> "sage-topology". And if you don't consider graphs as a graphs or
> topological spaces, then they appear to be combinatorial objects...

I never claimed that graph theory was not a subfield of combinatorics.
What I fight against is this group's pretense to control everything
that is connected with combinatorics, by 'coining' a general
mathematical term. Sage-combinat is a group headed by people who use
"combinat" code to request funding for their own ends, and I want to
make it very very clear that whoever claims that the work I put into
Sage's graph library is part of sage-combinat will see me claim the
opposite, every single time. By personnal letters sent to the funding
agencies if needed.

I already took it very very bad to see in the H2020 proposal that
"Paris-Sud" was one of the main centers for Sage development, where
basically nobody but me contributes in this area (and I was not part
of this proposal).

Play it fairly.

Nathann

[Nicolas M. Thiery]

On Mon, Jun 22, 2015 at 01:24:24PM +0200, Nathann Cohen wrote:
> I never claimed that graph theory was not a subfield of combinatorics.
> What I fight against is this group's pretense to control everything
> that is connected with combinatorics, by 'coining' a general
> mathematical term. Sage-combinat is a group headed by people who use
> "combinat" code to request funding for their own ends, and I want to
> make it very very clear that whoever claims that the work I put into
> Sage's graph library is part of sage-combinat will see me claim the
> opposite, every single time. By personnal letters sent to the funding
> agencies if needed.

Perfect: there is nothing to fight about then. Of course, when talking
to fellow researchers, I often advertised all the great features Sage
has that can be useful to people doing combinatorics; and this
includes graph theory, linear algebra, number theory, etc. But I don't
think anybody ever claimed that the graph theory code was part of the
"Sage-Combinat code". Well except you, about some chunks of the graph
theory code where I and a few others happened to implement features we
were missing :-)

In fact, even the notion of "Sage-Combinat code" is rather
meaningless. Sage-Combinat is not a library. It's just a community of
people that share some common interests in "improving Sage for
combinatorics ... and fostering code sharing in this area".

Early on, Sage-Combinat was a relatively well-defined small group of
people collaborating on the same foundations, which required tight
coordination. It has now grown to the point that it's progressively
dissolving into a nebula of subcommunities of Sage developers centered
on specific areas (Coxeter groups, crystals, symmetric functions,
enumerative combinatorics, ...). Drawing a line between those
communities and other subcommunities within Sage is becoming rather
meaningless.

That's great. In fact, it's all I have every wished for. When you
create an informal structure of some sort for a specific goal, the
best that can happen is when this structure becomes useless because
the main goal is being reached: it has become natural for many
researchers in combinatorics to get involved, share their code, and
collaborate.

Some people, including me, have spent a great deal of time and energy
in the last 15 years for this to happen. Did this have an influence?
Or would this have happened anyway? That's not for me to judge. Yet I
believe it's fair for me (and others!) to report on this hard work in
applications to support the fact that, maybe, I have a bit of
experience in coordinating projects.

> I already took it very very bad to see in the H2020 proposal that
> "Paris-Sud" was one of the main centers for Sage development, where
> basically nobody but me contributes in this area (and I was not part
> of this proposal).

The proposal building was completely open and you have been welcome at
every step of it. You still are, if you wish. If the phrasing for the
description of Paris Sud did not suit you, you could have requested a
change. You could even have made a pull request.

That being said, the description of the local context in the proposal
seems rather suitable to me. It's just a fact that the University
Paris Sud physically hosts one of the largest groups of Sage
contributors, especially if you include people from nearby
institutions (not that it means so much given that the Sage community
is dispersed by nature). There is no question that you are currently
the most active dev at UPsud in terms of #tickets, etc). But it's not
being claimed that this activity has a priori any relation with the
H2020, except that it's at the same physical location.

Just two last notes:

- The H2020 has basically nothing to do with combinatorics (except for
one or two tasks).

- It's a lot about offering to some people the opportunity to devote
all their work time to improving Sage and sister projects. I agree,
it's only for four years. Permanent positions would be much better;
we will also work on this, but occasions are rare. As a small
compensation the salary is ok; as you mentioned elsewhere
potentially larger than yours. Potentially larger than mine, for
that matter.

Cheers,
Nicolas
--
Nicolas M. Thiéry "Isil" <nth...@users.sf.net>
http://Nicolas.Thiery.name/

[Nicolas M. Thiery]

On similar situations, we have made an overview track ticket (labelled
as "task"), with pointers to individual tickets. See e.g.:

#10673: Roadmap for (Combinatorial) Free Module

It's not necessarily a guarantee that there actually will be progress,
but at least it can help getting a bird eye view on what needs to be
done.

[Nicolas M. Thiery]

On Mon, Jun 22, 2015 at 01:09:57PM +0200, Nathann Cohen wrote:
> I was about to say that "such forums tend to restrict to a subset of
> developers some conversations that have an impact for everybody".
>
> I just loaded sage-combinat's forum, to find out that there is a
> question there about graphs, which of course are not part of combinat

The problem with sage-devel is that, for many people, there is way too
much traffic to keep track of it. Those thematic lists were an attempt
at solving this problem, but I agree it does not work that well.

Maybe an alternative would be to have a single list, but with some
tagging mechanism so that someone interested in some topics could just
follow the threads related to those topics.

As an intermediate step, we could make sure that every mail going to
e.g. sage-combinat-devel would be forwarded to sage-devel. For this
particular thread on graphs, I was actually about to forward it to
sage-devel; now that I know that you read sage-combinat-devel only
occasionally, I'll make sure to do it at once next time :-)