4.0 vs 3.8 again

classic Classic list List threaded Threaded
29 messages Options
12
Reply | Threaded
Open this post in threaded view
|

4.0 vs 3.8 again

Jordi Gutiérrez Hermoso-2
I know we've had this discussion before, but I want to revisit it
again. I know this is bikeshedding[1], so this will be my final word
on the matter, I promise. Or at least publicly, if anyone wants to
keep arguing this with me, we can do so while only the NSA watches.

I really think the next Octave release should be called "Octave 4".
Not "3.8".

I will restate my case, with a few extra bits of data that we now
have:

   1) Gauging by the demand for it as well as the development buzz
      that it has spurred, the GUI alone is gonna be *huge*. It's such
      a big feature that a change in the major revision is warranted,
      even if our other big features like JIT compiling or OpenGL are
      still betaish.

   2) It's good marketing. "Octave 4" is a simple, single number to
      remember. It's easy to convey to people which Octave version has
      the GUI, and which Octave version started to make a point of
      also making timely Windows and Mac OS X binary releases.

   3) The ".0" part of "4.0" gives an indication that we are trying
      something big and new here. Then 4.2 and beyond will start
      adding more polish, perhaps JIT or classdef, but for now the
      ".0" will express to our users that this is a major change. I
      hope that this will earn us some leeway and indulgence from our
      users if not everything is great, as long as the GUI is great.

   4) I would like to have a website redesign to go along with the new
      release, and a big "4" would look very nice in this redesign.

I won't be terribly disappointed if we do decide to go with "3.8", but
"Octave 4" just sounds so nice.

- Jordi G. H.

[1] http://en.wikipedia.org/wiki/Bikeshedding


Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

Thorsten Liebig
Am 19.11.2013 16:16, schrieb Jordi Gutiérrez Hermoso:

> I know we've had this discussion before, but I want to revisit it
> again. I know this is bikeshedding[1], so this will be my final word
> on the matter, I promise. Or at least publicly, if anyone wants to
> keep arguing this with me, we can do so while only the NSA watches.
>
> I really think the next Octave release should be called "Octave 4".
> Not "3.8".
>
> I will restate my case, with a few extra bits of data that we now
> have:
>
>    1) Gauging by the demand for it as well as the development buzz
>       that it has spurred, the GUI alone is gonna be *huge*. It's such
>       a big feature that a change in the major revision is warranted,
>       even if our other big features like JIT compiling or OpenGL are
>       still betaish.
>
>    2) It's good marketing. "Octave 4" is a simple, single number to
>       remember. It's easy to convey to people which Octave version has
>       the GUI, and which Octave version started to make a point of
>       also making timely Windows and Mac OS X binary releases.
>
>    3) The ".0" part of "4.0" gives an indication that we are trying
>       something big and new here. Then 4.2 and beyond will start
>       adding more polish, perhaps JIT or classdef, but for now the
>       ".0" will express to our users that this is a major change. I
>       hope that this will earn us some leeway and indulgence from our
>       users if not everything is great, as long as the GUI is great.
>
>    4) I would like to have a website redesign to go along with the new
>       release, and a big "4" would look very nice in this redesign.
>
> I won't be terribly disappointed if we do decide to go with "3.8", but
> "Octave 4" just sounds so nice.
>
> - Jordi G. H.
>
> [1] http://en.wikipedia.org/wiki/Bikeshedding
>
>

I completely agree and never understood why this should be just another 3.x in the first place...

What should happen for a 4.0 if not the changes made NOW..

regards
Thorsten

Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

Robert T. Short
In reply to this post by Jordi Gutiérrez Hermoso-2
On 11/19/2013 07:16 AM, Jordi Gutiérrez Hermoso wrote:

> I know we've had this discussion before, but I want to revisit it
> again. I know this is bikeshedding[1], so this will be my final word
> on the matter, I promise. Or at least publicly, if anyone wants to
> keep arguing this with me, we can do so while only the NSA watches.
>
> I really think the next Octave release should be called "Octave 4".
> Not "3.8".
>
> I will restate my case, with a few extra bits of data that we now
> have:
>
>     1) Gauging by the demand for it as well as the development buzz
>        that it has spurred, the GUI alone is gonna be *huge*. It's such
>        a big feature that a change in the major revision is warranted,
>        even if our other big features like JIT compiling or OpenGL are
>        still betaish.
>
>     2) It's good marketing. "Octave 4" is a simple, single number to
>        remember. It's easy to convey to people which Octave version has
>        the GUI, and which Octave version started to make a point of
>        also making timely Windows and Mac OS X binary releases.
>
>     3) The ".0" part of "4.0" gives an indication that we are trying
>        something big and new here. Then 4.2 and beyond will start
>        adding more polish, perhaps JIT or classdef, but for now the
>        ".0" will express to our users that this is a major change. I
>        hope that this will earn us some leeway and indulgence from our
>        users if not everything is great, as long as the GUI is great.
>
>     4) I would like to have a website redesign to go along with the new
>        release, and a big "4" would look very nice in this redesign.
>
> I won't be terribly disappointed if we do decide to go with "3.8", but
> "Octave 4" just sounds so nice.
>
> - Jordi G. H.
>
> [1] http://en.wikipedia.org/wiki/Bikeshedding
>
>
>
>

Jordi is right - a release with major feature additions probably should
get a major release bump (although this is contraindicated if we were
using semantic versioning) .  But how stable is the "GUI"? If it isn't
stable, the major bump should wait.  Judging from the traffic on this
list, I would say wait.  The other really major feature is classdef
support, but I get the same sense there.  Should we allow for some
stability before making the bump?

Since I have never been able to build octave with the "GUI", I can't
judge this myself - I wouldn't use it anyway.

Bob
Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

c.-2
In reply to this post by Jordi Gutiérrez Hermoso-2

On 19 Nov 2013, at 16:16, Jordi Gutiérrez Hermoso <[hidden email]> wrote:

> I know we've had this discussion before, but I want to revisit it
> again. I know this is bikeshedding[1], so this will be my final word
> on the matter, I promise. Or at least publicly, if anyone wants to
> keep arguing this with me, we can do so while only the NSA watches.
>
> I really think the next Octave release should be called "Octave 4".
> Not "3.8".
>
> I will restate my case, with a few extra bits of data that we now
> have:
>
>   1) Gauging by the demand for it as well as the development buzz
>      that it has spurred, the GUI alone is gonna be *huge*. It's such
>      a big feature that a change in the major revision is warranted,
>      even if our other big features like JIT compiling or OpenGL are
>      still betaish.
>
>   2) It's good marketing. "Octave 4" is a simple, single number to
>      remember. It's easy to convey to people which Octave version has
>      the GUI, and which Octave version started to make a point of
>      also making timely Windows and Mac OS X binary releases.
>
>   3) The ".0" part of "4.0" gives an indication that we are trying
>      something big and new here. Then 4.2 and beyond will start
>      adding more polish, perhaps JIT or classdef, but for now the
>      ".0" will express to our users that this is a major change. I
>      hope that this will earn us some leeway and indulgence from our
>      users if not everything is great, as long as the GUI is great.
>
>   4) I would like to have a website redesign to go along with the new
>      release, and a big "4" would look very nice in this redesign.
>
> I won't be terribly disappointed if we do decide to go with "3.8", but
> "Octave 4" just sounds so nice.
>
> - Jordi G. H.
>
> [1] http://en.wikipedia.org/wiki/Bikeshedding


personally I don't think 3.8 vs 4.0 makes much difference, but I do
prefer 3.8 slightly over 4.0.

the main reason is that, although the GUI is a huge difference it is
still terribly buggy, at least on OSX where I am testing, I'd prefer
a new major release to be more stable.

I like the idea of the big 4 on the website,
just add "coming soon, stay tuned!" and it will create even more hype.

also this will give more sense to the news window and convince a few users
to leave it open and it's a good way to ask for donations, "4.0 is almost done, it will
be out sooner if you contribute!"


just my .02€

c.



Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

avlas

El Dimarts, 19 de novembre de 2013, a les 17:11:40, c. va escriure:

> On 19 Nov 2013, at 16:16, Jordi Gutiérrez Hermoso <[hidden email]>
wrote:

> > I know we've had this discussion before, but I want to revisit it
> > again. I know this is bikeshedding[1], so this will be my final word
> > on the matter, I promise. Or at least publicly, if anyone wants to
> > keep arguing this with me, we can do so while only the NSA watches.
> >
> > I really think the next Octave release should be called "Octave 4".
> > Not "3.8".
> >
> > I will restate my case, with a few extra bits of data that we now
> >
> > have:
> >   1) Gauging by the demand for it as well as the development buzz
> >  
> >      that it has spurred, the GUI alone is gonna be *huge*. It's such
> >      a big feature that a change in the major revision is warranted,
> >      even if our other big features like JIT compiling or OpenGL are
> >      still betaish.
> >  
> >   2) It's good marketing. "Octave 4" is a simple, single number to
> >  
> >      remember. It's easy to convey to people which Octave version has
> >      the GUI, and which Octave version started to make a point of
> >      also making timely Windows and Mac OS X binary releases.
> >  
> >   3) The ".0" part of "4.0" gives an indication that we are trying
> >  
> >      something big and new here. Then 4.2 and beyond will start
> >      adding more polish, perhaps JIT or classdef, but for now the
> >      ".0" will express to our users that this is a major change. I
> >      hope that this will earn us some leeway and indulgence from our
> >      users if not everything is great, as long as the GUI is great.
> >  
> >   4) I would like to have a website redesign to go along with the new
> >  
> >      release, and a big "4" would look very nice in this redesign.
> >
> > I won't be terribly disappointed if we do decide to go with "3.8", but
> > "Octave 4" just sounds so nice.
> >
> > - Jordi G. H.
> >
> > [1] http://en.wikipedia.org/wiki/Bikeshedding
>
> personally I don't think 3.8 vs 4.0 makes much difference, but I do
> prefer 3.8 slightly over 4.0.
>
> the main reason is that, although the GUI is a huge difference it is
> still terribly buggy, at least on OSX where I am testing, I'd prefer
> a new major release to be more stable.
>
> I like the idea of the big 4 on the website,
> just add "coming soon, stay tuned!" and it will create even more hype.
>
> also this will give more sense to the news window and convince a few users
> to leave it open and it's a good way to ask for donations, "4.0 is almost
> done, it will be out sooner if you contribute!"
>
>
> just my .02€
>
> c.

From a user point of view, 4.0 is definitely more appealing... said that, I
recall all the criticism kde received when it buggyly transitioned to 4.0, and
then it took some time before good reviews came back

Why don't invest energy on getting stable gui, jit, etc and then release it?
This could be an option, unless it would take not few months but ~1-2 years to
make this really happen...

Again, only from a user point of view...

Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

c.-2
In reply to this post by c.-2

On 19 Nov 2013, at 17:11, c. <[hidden email]> wrote:

> the main reason is that, although the GUI is a huge difference it is
> still terribly buggy, at least on OSX where I am testing, I'd prefer
> a new major release to be more stable.

BTW, although I'm mostly looking at the GUI because, as you say,
that is a HUGE new feature, problems on OSX 10.9 are not only there,
I still see ~15 test failures in 'make check'.

c.
Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

John W. Eaton
Administrator
In reply to this post by avlas
On 11/19/2013 11:37 AM, Salva Ardid wrote:

> Why don't invest energy on getting stable gui, jit, etc and then release it?
> This could be an option, unless it would take not few months but ~1-2 years to
> make this really happen...

I'm pretty sure I can guarantee that it's going to take more than a few
months for the JIT compiler to be working and doing anything significant.

And for "etc", I guess it depends on what you mean by "etc".

jwe


Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

avlas


El Dimarts, 19 de novembre de 2013, a les 12:16:01, John W. Eaton va escriure:

> On 11/19/2013 11:37 AM, Salva Ardid wrote:
> > Why don't invest energy on getting stable gui, jit, etc and then release
> > it? This could be an option, unless it would take not few months but ~1-2
> > years to make this really happen...
>
> I'm pretty sure I can guarantee that it's going to take more than a few
> months for the JIT compiler to be working and doing anything significant.
>
> And for "etc", I guess it depends on what you mean by "etc".
>
> jwe

'etc' was referred to classdef as well as any other significant improvement
that I may not be aware of, which is planned to be included in GNU Octave 3.8
(or 4.0)...

Perhaps, deciding on 4.0 might be based on what is the main point you want to
really emphasize with it: is it the GUI? is it JIT? any other? and then see
when this may be ready for the average user (to avoid good expectations to be
counterproductive). Finally, according to that you could decide whether next
version can be better call 3.8 or 4.0, and the most proper time for it to
appear...

Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

c.-2
In reply to this post by c.-2

On 19 Nov 2013, at 17:58, c. <[hidden email]> wrote:

> BTW, although I'm mostly looking at the GUI because, as you say,
> that is a HUGE new feature, problems on OSX 10.9 are not only there,
> I still see ~15 test failures in 'make check'.

and, you know, in the old days when I used to do VLSI design I was taught
you should never release anything if you are not absolutely positive that
it is 100% bug free :D

c.
Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

Jordi Gutiérrez Hermoso-2
In reply to this post by c.-2
On Tue, 2013-11-19 at 17:58 +0100, c. wrote:
> On 19 Nov 2013, at 17:11, c. <[hidden email]> wrote:
>
> > the main reason is that, although the GUI is a huge difference it is
> > still terribly buggy, at least on OSX where I am testing, I'd prefer
> > a new major release to be more stable.
>
> BTW, although I'm mostly looking at the GUI because, as you say,
> that is a HUGE new feature, problems on OSX 10.9 are not only there,
> I still see ~15 test failures in 'make check'.

We should consider these blocker bugs.

If we can't make this release with the GUI working on Mac OS X and
Windows, then we shouldn't make it at all. The GUI should be our major
release goal.

I'm doing a Octave builds on Mac OS X.VIII and X.V, and I'll start
attacking these bugs.

- Jordi G. H.


Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

c.-2

On 19 Nov 2013, at 18:32, Jordi Gutiérrez Hermoso <[hidden email]> wrote:

> Mac OS X.VIII and X.V

I am afraid that is not going to help much, I did not get those
failures on OSX 10.8, and Ben doesn't get them on 10.7 so I guess
these are specific to 10.9.
Maybe you should also upgrade to 10.9 before building?

how does octave-gui work in your build? if you are also working on OSX,
I  could really use help in solving this bug (for OSX 10.8):
https://savannah.gnu.org/bugs/?40498

or this one (for OSX 10.9)
https://savannah.gnu.org/bugs/?40545

c.
Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

Doug Stewart-4
In reply to this post by Jordi Gutiérrez Hermoso-2



On Tue, Nov 19, 2013 at 12:32 PM, Jordi Gutiérrez Hermoso <[hidden email]> wrote:
On Tue, 2013-11-19 at 17:58 +0100, c. wrote:
> On 19 Nov 2013, at 17:11, c. <[hidden email]> wrote:
>
> > the main reason is that, although the GUI is a huge difference it is
> > still terribly buggy, at least on OSX where I am testing, I'd prefer
> > a new major release to be more stable.
>
> BTW, although I'm mostly looking at the GUI because, as you say,
> that is a HUGE new feature, problems on OSX 10.9 are not only there,
> I still see ~15 test failures in 'make check'.

We should consider these blocker bugs.

If we can't make this release with the GUI working on Mac OS X and
Windows, then we shouldn't make it at all. The GUI should be our major
release goal.

I'm doing a Octave builds on Mac OS X.VIII and X.V, and I'll start
attacking these bugs.

- Jordi G. H.



In my opinion Octave 4 has to be stable in windows and mack before we have the BIG release.

 I would like to see a release candidate out soon, because even on Ubuntu 12.04 I can't get a compile that works correctly. ( can't open an old .m file with the GUI)
 
Doug

--
DAS

Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

c.-2
In reply to this post by Jordi Gutiérrez Hermoso-2

On 19 Nov 2013, at 18:32, Jordi Gutiérrez Hermoso <[hidden email]> wrote:

> We should consider these blocker bugs.

I'd rather not.

It might take quite some time before they are all fixed and, IMHO,
we really need to get the GUI out for testing by a larger audience.

Why not release 3.8 now and leave 4.0 for when these bugs are fixed?

c.


Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

Michael Godfrey
In reply to this post by c.-2
On 11/19/2013 11:58 AM, c. wrote:
On 19 Nov 2013, at 17:11, c. [hidden email] wrote:

> the main reason is that, although the GUI is a huge difference it is 
> still terribly buggy, at least on OSX where I am testing, I'd prefer 
> a new major release to be more stable.
BTW, although I'm mostly looking at the GUI because, as you say, 
that is a HUGE new feature, problems on OSX 10.9 are not only there,
I still see ~15 test failures in 'make check'.

c.
Is it worth considering a 3.8 release without the GUI, and a 4.0 release with
GUI when the GUI is definitely suitable for general use on all platforms?
This may satisfy the differing views about 3.8 and 4.0 and avoid releasing
a buggy GUI.

Michael

Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

John Swensen-3
In reply to this post by c.-2
On Tue, Nov 19, 2013 at 12:42 PM, c. <[hidden email]> wrote:

On 19 Nov 2013, at 18:32, Jordi Gutiérrez Hermoso <[hidden email]> wrote:

> Mac OS X.VIII and X.V

I am afraid that is not going to help much, I did not get those
failures on OSX 10.8, and Ben doesn't get them on 10.7 so I guess
these are specific to 10.9.
Maybe you should also upgrade to 10.9 before building?

how does octave-gui work in your build? if you are also working on OSX,
I  could really use help in solving this bug (for OSX 10.8):
https://savannah.gnu.org/bugs/?40498

or this one (for OSX 10.9)
https://savannah.gnu.org/bugs/?40545

c.

I have run the Octave.app with the GUI from a couple of weeks back on 10.9 with no problems.  I haven't done a lot of work with it, but have done some simple scripts.
Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

c.-2

On 19 Nov 2013, at 19:39, John Swensen <[hidden email]> wrote:

> I have run the Octave.app with the GUI from a couple of weeks back on 10.9 with no problems.  I haven't done a lot of work with it, but have done some simple scripts.

I made that build, that is linked to qscintilla 2.6 so it does not have syntax highlighting in the editor.
also it cannot install packages.

c.
Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

c.-2
In reply to this post by Michael Godfrey

On 19 Nov 2013, at 19:37, Michael D. Godfrey <[hidden email]> wrote:

> Is it worth considering a 3.8 release without the GUI, and a 4.0 release with
> GUI when the GUI is definitely suitable for general use on all platforms?
> This may satisfy the differing views about 3.8 and 4.0 and avoid releasing
> a buggy GUI.
>
> Michael


or maybe also ship the GUI with 3.8 but
have octave be octave-cli for 3.8
and octave-gui for 4.0?

c.
Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

Michael Godfrey
On 11/19/2013 01:56 PM, c. wrote:
On 19 Nov 2013, at 19:37, Michael D. Godfrey [hidden email] wrote:

> Is it worth considering a 3.8 release without the GUI, and a 4.0 release with
> GUI when the GUI is definitely suitable for general use on all platforms?
> This may satisfy the differing views about 3.8 and 4.0 and avoid releasing
> a buggy GUI.
> 
> Michael
or maybe also ship the GUI with 3.8 but
have octave be octave-cli for 3.8
and octave-gui for 4.0?

c.
Right.  I really meant to say 3.8 would be default CLI, but allow experimental GUI.


Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

Rik-4
In reply to this post by Jordi Gutiérrez Hermoso-2
11/19/13

All,

I'm not pathologically attached to either 3.8 or 4.0.  I suggested 3.8
early on because 1) it has been almost two years (Sun Jan 15 13:02:42 2012)
since the 3.6.X series started and I don't want to keep waiting for the
code to come out, and 2) the finished GUI (stable, Ctrl+C interrupts work,
Qt instead of FLTK widgets, all platforms working) still seems a ways off.
I think the GUI is the number one feature desired (forget JIT, even
classdef), and it was described to me that incorporating Qt plotting and
the cleanup work might take 6 months.

Right now the GUI is imperfect, for example there is an architectural flaw
which is preventing FLTK plotting on Mac OSX platforms.  This is probably
never going to be fixed, rather it will be resolved by using the Qt widget
set which we want to do anyways and which uses a different software
architecture.  So it seems appropriate to release this GUI under a 3.8
labeling with the full expectation that is is going to be superseded and
quite different in a 4.0 release.  You could do this as well with a 4.0 to
4.2 transition, but I think people will have higher expectations for a 4.0
release than they will for one which is a continuation of a series.

--Rik
Reply | Threaded
Open this post in threaded view
|

Re: 4.0 vs 3.8 again

Michael Godfrey
On 11/19/2013 02:42 PM, Rik wrote:

> 11/19/13
>
> All,
>
> I'm not pathologically attached to either 3.8 or 4.0.  I suggested 3.8
> early on because 1) it has been almost two years (Sun Jan 15 13:02:42 2012)
> since the 3.6.X series started and I don't want to keep waiting for the
> code to come out, and 2) the finished GUI (stable, Ctrl+C interrupts work,
> Qt instead of FLTK widgets, all platforms working) still seems a ways off.
> I think the GUI is the number one feature desired (forget JIT, even
> classdef), and it was described to me that incorporating Qt plotting and
> the cleanup work might take 6 months.
>
> Right now the GUI is imperfect, for example there is an architectural flaw
> which is preventing FLTK plotting on Mac OSX platforms.  This is probably
> never going to be fixed, rather it will be resolved by using the Qt widget
> set which we want to do anyways and which uses a different software
> architecture.  So it seems appropriate to release this GUI under a 3.8
> labeling with the full expectation that is is going to be superseded and
> quite different in a 4.0 release.  You could do this as well with a 4.0 to
> 4.2 transition, but I think people will have higher expectations for a 4.0
> release than they will for one which is a continuation of a series.
>
> --Rik
This seems right to me.  3.8 then 4.0 seems reasonable, but 4.0 then 4.2
is fine if it is preferred by others.  The key thing is to get a stable
release out very
soon.

Michael

12