plplot-devel

[Plplot-devel] Where do we go from here?

[Alan W. I. <ir...@be...>]

I understand that Geoffrey is still preparing his state of the union message
(or speech from the throne here in Canada.)  But in light of my semi-humorous
remark about the next release in my last post, I decided to jump start that
discussion.

My preference for the next release is I would like to see a continued effort
at the small bug fixes, documentation, example consistency, Java API, etc.,
mixed with at least one major improvement. Furthermore, I believe that major
improvement should be an AM/LT configuration scheme.  It did not escape my
notice that a substantial fraction of our commits since 5.0.4 had to do with
configuration changes, and I think that ratio would be greatly reduced with
AM/LT once we had overcome the learning curve.  I wouldn't feel comfortable
leading the AM/LT project, but once Rafael (or some other knowledgeable
volunteer) had time to lead that effort, then I would be willing to spend
some time on a CVS branch helping him with testing the new scheme. Once my
AM/LT leader and I were satisfied, I would hope the rest of the developers
would give the AM/LT configuration a try so that we were all comfortable
bringing it to CVS head for final refinement.

So I am looking for an AM/LT leader I could follow.  Rafael would be ideal
for this role (and we have worked well together before), but he may not be
able to make a time commitment in the near future so we may have to rely on
somebody else.

Alan

email: ir...@be...
phone: 250-727-2902	FAX: 250-721-7715
snail-mail:
Dr. Alan W. Irwin
Department of Physics and Astronomy,
University of Victoria, P.O. Box 3055,
Victoria, British Columbia, Canada, V8W 3P6
__________________________

Linux-powered astrophysics
__________________________

RE: [Plplot-devel] Where do we go from here?

[Geoffrey F. <fu...@ga...>]

Alan W. Irwin writes:
 > I understand that Geoffrey is still preparing his state of the union message

Sorry, I know I said I'd already have transmitted it by now, but the
last few days hae gone wacky on me.  

Will try to push it out "real soon now".

-- 
Geoffrey Furnish         fu...@ga...

Re: [Plplot-devel] Where do we go from here?

[Rafael L. <lab...@mp...>]

* Maurice LeBrun <mj...@ga...> [2002-02-06 21:56]:

> I'm a bit skeptical it would have made a big difference.  It seems to me
> that with the addition of dyndrivers, there were going to be a lot of
> configuration changes regardless of which configuration model we were
> using.

I do not remember if I told you guys about this, but in my private AM/LT
branch (not in CVS) I started porting the dyndrv code to libltdl, a portable
library for manipulating dynamically loaded objects (see
http://www.gnu.org/manual/libtool-1.4.2/html_node/libtool_45.html).  The
goal is to make the dyndrv code largely portable across platforms. By no
means, using libltdl means using libtool, but the two things work nice
together.

> Certainly the commits I made personally left me with that impression.  I
> don't mean to be throwing water on your enthusiasm, I just think it will
> take a while for the payoff from AM/LT (or dyndrivers for that matter) to
> be really evident.  The payoff from cool new features, however, can be
> immediate.

I also think that AM/LT will not bring anything visible for the end-user. My
interest in AM/LT is from the packager point of view.  Believe me, the
current configuration can be a nightmare for people packaging PLplot
(witness rpath problems, non-relocatability, etc).  Besides that,
integrating the TEA stuff may be an easier task in the framework of AM/LT.
I also think that making the configuration scheme more "modern" (yeah, in
the GNU sense...) will make it much more attractive for the potential future
developers joining our team.

> Yeah, I'd feel the most comfortable if the effort were led by Rafael.
> Hopefully no time real soon so I can get some other work done. :)

No doubt, I am the most entitled person to do the job.  Unfortunately, I
have this time budget problem right now.  At any rate, I did not resign from
the project yet, but if I see that I cannot do anything to improve the
situation, I will have to make a decision.  I will tell you when/if this
arrive (I hope not).

-- 
Rafael

Re: [Plplot-devel] Where do we go from here?

[Alan W. I. <ir...@be...>]

On Thu, 7 Feb 2002, Rafael Laboissiere wrote:

> [...] Unfortunately, I
> have this time budget problem right now.  At any rate, I did not resign from
> the project yet, but if I see that I cannot do anything to improve the
> situation, I will have to make a decision.  I will tell you when/if this
> arrive (I hope not).

I presume you are talking about just the AM/LT project, Rafael. Your total
contributions to PLplot have been most valuable, and I hope they long
continue on whatever sub-project you decide to work on.  Also, note there is
no hurry on the AM/LT decision.  Did you notice all the rush of volunteers
that I got?  ...;-)

The only reason I am agitating about AM/LT now is I feel it is important
infrastructure we should at least test preferably at the start of *a*
release cycle.  If it turns out we don't get it into the release in this
cycle, I will definitely be asking about AM/LT again near the start of the
next release cycle since I do not want to lose all the valuable work you put
into the AM/LT branch.

Alan

Re: [Plplot-devel] Where do we go from here?

[Maurice L. <mj...@ga...>]

Alan W. Irwin writes:
 > Furthermore, I believe that major
 > improvement should be an AM/LT configuration scheme.  It did not escape my
 > notice that a substantial fraction of our commits since 5.0.4 had to do with
 > configuration changes, and I think that ratio would be greatly reduced with
 > AM/LT once we had overcome the learning curve.

I'm a bit skeptical it would have made a big difference.  It seems to me that
with the addition of dyndrivers, there were going to be a lot of configuration
changes regardless of which configuration model we were using.  Certainly the
commits I made personally left me with that impression.  I don't mean to be
throwing water on your enthusiasm, I just think it will take a while for the
payoff from AM/LT (or dyndrivers for that matter) to be really evident.  The
payoff from cool new features, however, can be immediate.

 > So I am looking for an AM/LT leader I could follow.  Rafael would be ideal
 > for this role (and we have worked well together before), but he may not be
 > able to make a time commitment in the near future so we may have to rely on
 > somebody else.

Yeah, I'd feel the most comfortable if the effort were led by Rafael.
Hopefully no time real soon so I can get some other work done. :)

-- 
Maurice LeBrun    mj...@ga...

Re: [Plplot-devel] Where do we go from here?

[Alan W. I. <ir...@be...>]

> Alan W. Irwin writes:
>  > Furthermore, I believe that major
>  > improvement should be an AM/LT configuration scheme.  It did not escape my
>  > notice that a substantial fraction of our commits since 5.0.4 had to do with
>  > configuration changes, and I think that ratio would be greatly reduced with
>  > AM/LT once we had overcome the learning curve.
>
> I'm a bit skeptical it would have made a big difference.  It seems to me that
> with the addition of dyndrivers, there were going to be a lot of configuration
> changes regardless of which configuration model we were using.  Certainly the
> commits I made personally left me with that impression.  I don't mean to be
> throwing water on your enthusiasm, I just think it will take a while for the
> payoff from AM/LT (or dyndrivers for that matter) to be really evident.  The
> payoff from cool new features, however, can be immediate.

IMHO, if we improve the configuration organization and overall
infrastructure, then it will make the interesting "cool new features"
substantially easier to do (e.g., fewer and easier configuration commits
then what occurred with the dyndriver changes.) I am glad we continue to
desuckify the current configuration scheme (witness Geoffrey's comments on
the dependency problems that just bit me), but at some point you have to
evaluate how much time that is taking versus the advantages of a new
approach.

From the encouraging things I have read, I am willing to gamble with my time
doing testing on the AM/LT branch that automake and libtool will make our
life considerably easier.  But I don't want to finish Rafael's experiment
alone.

Alan

Re: [Plplot-devel] Where do we go from here?

[Geoffrey F. <fu...@ga...>]

Rafael Laboissiere writes:
Quoted out of order:
 > > Yeah, I'd feel the most comfortable if the effort were led by Rafael.
 > > Hopefully no time real soon so I can get some other work done. :)
 > 
 > No doubt, I am the most entitled person to do the job.  Unfortunately, I
 > have this time budget problem right now.  At any rate, I did not resign from
 > the project yet, but if I see that I cannot do anything to improve the
 > situation, I will have to make a decision.  I will tell you when/if this
 > arrive (I hope not).

Hey buddy, you can run but you can't hide.  Once a member always a
member.  :-).  Seriously, we all face differing time demands,
diverging personal sw and research interests, etc.  If you need to
spend N months doing something else, even an indeterminately long
dive, that's no reason to feel you need to "resign".  Maurice and I
have both had long fallow periods on PLplot, whilst pushing other
agendas. Don't sweat it.  We're happy to have your constructive
contributions whenever you have the time.  And we'll push on as best
we can the rest of the time.  Really, its that way for all of us.  My
personal ambitions for what I wish I could accomplish on PLplot /far/
exceed what I can actually deliver, measured over any human
timescale of interest.  Such is life.

 > * Maurice LeBrun <mj...@ga...> [2002-02-06 21:56]:
 > 
 > > I'm a bit skeptical it would have made a big difference.  It seems to me
 > > that with the addition of dyndrivers, there were going to be a lot of
 > > configuration changes regardless of which configuration model we were
 > > using.
 > 
 > I do not remember if I told you guys about this, but in my private AM/LT
 > branch (not in CVS) I started porting the dyndrv code to libltdl, a portable
 > library for manipulating dynamically loaded objects (see
 > http://www.gnu.org/manual/libtool-1.4.2/html_node/libtool_45.html).  The
 > goal is to make the dyndrv code largely portable across platforms. By no
 > means, using libltdl means using libtool, but the two things work nice
 > together.
 > 
 > > Certainly the commits I made personally left me with that impression.  I
 > > don't mean to be throwing water on your enthusiasm, I just think it will
 > > take a while for the payoff from AM/LT (or dyndrivers for that matter) to
 > > be really evident.  The payoff from cool new features, however, can be
 > > immediate.
 > 
 > I also think that AM/LT will not bring anything visible for the end-user. My
 > interest in AM/LT is from the packager point of view.  Believe me, the
 > current configuration can be a nightmare for people packaging PLplot
 > (witness rpath problems, non-relocatability, etc).  Besides that,
 > integrating the TEA stuff may be an easier task in the framework of AM/LT.
 > I also think that making the configuration scheme more "modern" (yeah, in
 > the GNU sense...) will make it much more attractive for the potential future
 > developers joining our team.

I'll say again that I share Maurice's (healthy) skepticism, but both
of Rafael's pp's above have interesting, even tantalizing aspects, so
we should by all means endeavor to carry this through to the point
where it can be seriously evaluated.

My only real beef wit the "modern GNU (autoconf) way", is that it is
so woefully inadequate for the overwhelming majority of the software
development I do.  Which is to say, day-in-and-day-out development of
highly levelized C++ software.  Okay, those words don't describe
PLplot much, so maybe a more "modern" autoconf scheme would be okay
for PLplot.  I'm willing to be shown.  But I've just never really felt
that the GNU autoconf system really brings much to the table in the
area of software organization, layout, and general developer-centric
sw issues.  What it does wonderfully, par-excellance, is system
suckiness deobfuscation and workaround demystification.  And that's
pretty much what we use it for, to my mind, with substantial benefit.
And if I could read Maurice's mind from here, I'm guessing he's
feeling about the same, that we are deriving the main benefit of
autoconf right now.  In fact, we are actually deriving more than the
usual benefit, because Maurice has spent so much time (way back at the
beginning) desuckifying some of the autoconf macros.  Maybe they're
catching up by now, but at least for several years, our autoconf stuff
worked a lot better than the average.  I cannot count the number of
times I was able to configure PLplot, but not other GNU software, on
various unix systems where the swadmins had done various cockamaymee
nonsense.  This was especially bad in the labs, where "national
security" was such an oft-abused rubric for excusing any imaginable
sort of sysadmin-lunacy.  X headers under /dont/look/here, X libs
under /you/wont/believe/its/over/here, etc.  

I dunno, we need to see it working.  It may well be worth the switch.
I think the summary for Maurice and me, is just "let's not decide
before we see the proof".  But by all means, if the evidence can be
seen, we'll look.

-- 
Geoffrey Furnish         fu...@ga...

Re: [Plplot-devel] Where do we go from here?

[Rafael L. <lab...@mp...>]

* Geoffrey Furnish <fu...@ga...> [2002-02-07 10:27]:

> Hey buddy, you can run but you can't hide.  Once a member always a member.
> :-).  Seriously, we all face differing time demands, diverging personal sw
> and research interests, etc.  If you need to spend N months doing something
> else, even an indeterminately long dive, that's no reason to feel you need
> to "resign".  Maurice and I have both had long fallow periods on PLplot,
> whilst pushing other agendas. Don't sweat it.  We're happy to have your
> constructive contributions whenever you have the time.  And we'll push on
> as best we can the rest of the time.  Really, its that way for all of us.
> My personal ambitions for what I wish I could accomplish on PLplot /far/
> exceed what I can actually deliver, measured over any human timescale of
> interest.  Such is life.

* Alan W. Irwin <ir...@be...> [2002-02-07 08:36]:

> On Thu, 7 Feb 2002, Rafael Laboissiere wrote:
> 
> > [...] Unfortunately, I
> > have this time budget problem right now.  At any rate, I did not resign from
> > the project yet, but if I see that I cannot do anything to improve the
> > situation, I will have to make a decision.  I will tell you when/if this
> > arrive (I hope not).
> 
> I presume you are talking about just the AM/LT project, Rafael. Your total
> contributions to PLplot have been most valuable, and I hope they long 
> continue on whatever sub-project you decide to work on.  Also, note there
> is no hurry on the AM/LT decision. 

First of all, thanks for your kind words.  I have to apologize for the
extremely bad wording in the paragraph above.  When I wrote "I will have to
make a decision" I was not meaning resignation from the PLplot project, but
as Alan correctly inferred, I was referring to the AM/LT front.  I am still
willing to lead the work on this front, but if I cannot resume it in the
foreseeable future, it is better that someone else take it over from me.

-- 
Rafael