www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Is D releasing too often?

reply Joakim <dlang joakim.fea.st> writes:
There have been 6 major releases of dmd over the last year, with 
ldc trying to keep pace, currently only one release behind. This 
is a big jump up from the previous release schedule, I see 2 
major releases in 2014, 3 in 2015, and 3 in 2016.

There are obviously pros and cons to each pace, and this has been 
debated internally before, with one of the ldc devs again posting 
to the Internals mailing list today questioning the current speed.

I thought I'd open it up to the community: now that you've 
experienced this faster pace, as a user of the D compilers, how 
do you like it? Would you prefer a slower release schedule, say 
3-4 major releases a year?

I thought 6/year was an ambitious schedule when announced and I 
wonder if it isn't putting too much strain on our few release 
maintainers, maybe 3-4 releases/year would be a more gradual bump 
up.
May 14 2018
next sibling parent reply Mike Franklin <slavo5150 yahoo.com> writes:
On Monday, 14 May 2018 at 07:20:48 UTC, Joakim wrote:


 There are obviously pros and cons to each pace, and this has 
 been debated internally before, with one of the ldc devs again 
 posting to the Internals mailing list today questioning the 
 current speed.


The post is here: 
https://forum.dlang.org/post/ingkexhebiaqdzknicvr forum.dlang.org

My interpretation of that post is it's not the frequency of 
releases that's the problem, but rather that bug fixes and 
language changes are mixed together in each release.  If it's 
really a problem, one potential solution could be to keep bug fix 
releases every 2 months, but only release language changes every 
4 or 6 months.  But that's not cost-free either, as it will put 
more of a burden on compiler developers and release managers to 
maintain parallel branches and deal with the inevitable hassle 
and controversy involved in deciding which PR goes in which 
branch, as sometimes bug fixes and language changes are the same 
thing.  Reviewing PRs requires quite a bit of subjective 
judgement and conflicting opinions, and it's best to not add more 
aggravation to an already aggravating process.


 I thought I'd open it up to the community: now that you've 
 experienced this faster pace, as a user of the D compilers, how 
 do you like it? Would you prefer a slower release schedule, say 
 3-4 major releases a year?


I think the pace is just about right, though I could even 
tolerate even more frequent releases.  Less frequent releases 
would actually be a somewhat irritating to me.  For example, I 
have a language change in 2.081 that I really need in LDC today, 
but it looks like I'm going to have to wait at least 4 months to 
get it.

IMO, the current schedule is working out quite well.

Mike
May 14 2018
parent Uknown <sireeshkodali1 gmail.com> writes:
On Monday, 14 May 2018 at 07:46:35 UTC, Mike Franklin wrote:

 On Monday, 14 May 2018 at 07:20:48 UTC, Joakim wrote:


 [...]


 The post is here: 
 https://forum.dlang.org/post/ingkexhebiaqdzknicvr forum.dlang.org
 [...]
 Mike


I agree with you. I think the best way is to make it one release 
every 3 months or so and have a yearly LTS release that people 
can rely on for stability. On the plus side, I really like the 
predictable realeases.
May 14 2018
prev sibling next sibling parent KingJoffrey <KingJoffrey KingJoffrey.com> writes:
On Monday, 14 May 2018 at 07:20:48 UTC, Joakim wrote:

 I thought 6/year was an ambitious schedule when announced and I 
 wonder if it isn't putting too much strain on our few release 
 maintainers, maybe 3-4 releases/year would be a more gradual 
 bump up.


This is what happens to programming langauges and software 
development, when competition becomes the main game.

Thank god for Redhat - still on gcc 4.8.5 .. god bless them ;-)
May 14 2018
prev sibling next sibling parent reply Jonathan M Davis <newsgroup.d jmdavisprog.com> writes:
On Monday, May 14, 2018 07:20:48 Joakim via Digitalmars-d wrote:

 There have been 6 major releases of dmd over the last year, with
 ldc trying to keep pace, currently only one release behind. This
 is a big jump up from the previous release schedule, I see 2
 major releases in 2014, 3 in 2015, and 3 in 2016.

 There are obviously pros and cons to each pace, and this has been
 debated internally before, with one of the ldc devs again posting
 to the Internals mailing list today questioning the current speed.

 I thought I'd open it up to the community: now that you've
 experienced this faster pace, as a user of the D compilers, how
 do you like it? Would you prefer a slower release schedule, say
 3-4 major releases a year?

 I thought 6/year was an ambitious schedule when announced and I
 wonder if it isn't putting too much strain on our few release
 maintainers, maybe 3-4 releases/year would be a more gradual bump
 up.


I think that most of us are fine with how it's been going, but a large
percentage of the work for putting out each release goes to Martin, and he's
been talking about reducing the pace to something more like one major
release every 6 months, which I think would be too slow. I don't want to
have to wait that long for improvements to Phobos to become available.
Quarterly would be okay, I guess, but it's always frustrating when an
improvement makes it into Phobos but you can't use it for months, because
you have to wait for it to be released - and it's even worse when you want
to then be able to build your code with ldc, since they're always behind.

- Jonathan M Davis
May 14 2018
parent bauss <jj_1337 live.dk> writes:
On Monday, 14 May 2018 at 07:53:40 UTC, Jonathan M Davis wrote:

 On Monday, May 14, 2018 07:20:48 Joakim via Digitalmars-d wrote:

 There have been 6 major releases of dmd over the last year, 
 with ldc trying to keep pace, currently only one release 
 behind. This is a big jump up from the previous release 
 schedule, I see 2 major releases in 2014, 3 in 2015, and 3 in 
 2016.

 There are obviously pros and cons to each pace, and this has 
 been debated internally before, with one of the ldc devs again 
 posting to the Internals mailing list today questioning the 
 current speed.

 I thought I'd open it up to the community: now that you've 
 experienced this faster pace, as a user of the D compilers, 
 how do you like it? Would you prefer a slower release 
 schedule, say 3-4 major releases a year?

 I thought 6/year was an ambitious schedule when announced and 
 I wonder if it isn't putting too much strain on our few 
 release maintainers, maybe 3-4 releases/year would be a more 
 gradual bump up.


 I think that most of us are fine with how it's been going, but 
 a large percentage of the work for putting out each release 
 goes to Martin, and he's been talking about reducing the pace 
 to something more like one major release every 6 months, which 
 I think would be too slow. I don't want to have to wait that 
 long for improvements to Phobos to become available. Quarterly 
 would be okay, I guess, but it's always frustrating when an 
 improvement makes it into Phobos but you can't use it for 
 months, because you have to wait for it to be released - and 
 it's even worse when you want to then be able to build your 
 code with ldc, since they're always behind.

 - Jonathan M Davis


Honestly Phobos should be released independently from DMD (But 
should still be shipped with it of course.)

Library implementations shouldn't wait for the compiler to be 
available.

That's just my opinion though
May 14 2018
prev sibling next sibling parent reply "Nick Sabalausky (Abscissa)" <SeeWebsiteToContactMe semitwist.com> writes:
On 05/14/2018 03:20 AM, Joakim wrote:

 There have been 6 major releases of dmd over the last year, with ldc 
 trying to keep pace, currently only one release behind. This is a big 
 jump up from the previous release schedule, I see 2 major releases in 
 2014, 3 in 2015, and 3 in 2016.
 


This is just a return to the old original pace. It used to be about this 
often, but improved focus on avoiding regressions and packaging mistakes 
had a side effect of delaying releases down to as low as 2/year. But the 
improved "checks and balances and procedures" have finally become 
streamlined enough that the delays have been overcome and we're back to 
where we already were.

Personally, I like it. Keep in mind, availability of an update does not 
mandate immediate upgrading. It's just means its *available*. In an 
ideal world, improvements should be available immediately. In the real 
world, we have need to allow for both automated and manual checks for 
regressions. Therefore, the only things that should delay deployment of 
a project's new releases are:

A. Automated regression tests (and fixing found regressions).
B. A formal "beta" cycle (and fixing found regressions).
C. The existence of improvements (without improvements, what's the point 
of a new release?).

If dependent projects/libs have trouble keeping up with an increased 
rate of releases, that only means that we need better tools for 
automating the tasks involved in supporting newer releases.

If there's motivation to delay releases, that's a clear indication that 
dependent projects are in simply need of better automation/processes/etc.

Summary (That's "TL:DR" in hipster parlance): **Available** updates are 
always a good thing, as long as they're checked for regressions. Any 
issues keeping up with updates only indicate we have a *good* problem: 
That the ecosystem has room for "process" improvements and isn't 
currently being constrained by lack of progress in dependencies.
May 14 2018
next sibling parent reply Uknown <sireeshkodali1 gmail.com> writes:
On Monday, 14 May 2018 at 08:29:30 UTC, Nick Sabalausky 
(Abscissa) wrote:

 On 05/14/2018 03:20 AM, Joakim wrote:

 [..]


 Summary (That's "TL:DR" in hipster parlance): **Available** 
 updates are always a good thing, as long as they're checked for 
 regressions. Any issues keeping up with updates only indicate 
 we have a *good* problem: That the ecosystem has room for 
 "process" improvements and isn't currently being constrained by 
 lack of progress in dependencies.


Agree, but I think LTS releases are worth considering. A yearly 
release that gets only bug fixes would be useful for stability.
May 14 2018
parent "Nick Sabalausky (Abscissa)" <SeeWebsiteToContactMe semitwist.com> writes:
On 05/14/2018 04:35 AM, Uknown wrote:

 
 Agree, but I think LTS releases are worth considering. A yearly release 
 that gets only bug fixes would be useful for stability.


A fair point. I have no disagreement with something like that.
May 14 2018
prev sibling parent "Nick Sabalausky (Abscissa)" <SeeWebsiteToContactMe semitwist.com> writes:
Ie, as long as there's no regression problems: If the rate of releases 
makes us feel like we have an embarrassment of riches, it doesn't mean 
that we *NEED* fewer releases. It means we need to be more appreciative 
that there's all those enhancements we don't have to wait for.
May 14 2018
prev sibling next sibling parent Dejan Lekic <dejan.lekic gmail.com> writes:
On Monday, 14 May 2018 at 07:20:48 UTC, Joakim wrote:

 There have been 6 major releases of dmd over the last year, 
 with ldc trying to keep pace, currently only one release 
 behind. This is a big jump up from the previous release 
 schedule, I see 2 major releases in 2014, 3 in 2015, and 3 in 
 2016.

 There are obviously pros and cons to each pace, and this has 
 been debated internally before, with one of the ldc devs again 
 posting to the Internals mailing list today questioning the 
 current speed.

 I thought I'd open it up to the community: now that you've 
 experienced this faster pace, as a user of the D compilers, how 
 do you like it? Would you prefer a slower release schedule, say 
 3-4 major releases a year?

 I thought 6/year was an ambitious schedule when announced and I 
 wonder if it isn't putting too much strain on our few release 
 maintainers, maybe 3-4 releases/year would be a more gradual 
 bump up.


I have nothing against releasing that often as long as there is a 
LTS version, and that is where the problem lies - we do not have 
it. So what I do is simply base my production code on particular 
release, and every few months allocate few days job to try to 
bring the code up-to-date with latest DMD.
May 14 2018
prev sibling parent bachmeier <no spam.net> writes:
On Monday, 14 May 2018 at 07:20:48 UTC, Joakim wrote:


 I thought I'd open it up to the community: now that you've 
 experienced this faster pace, as a user of the D compilers, how 
 do you like it? Would you prefer a slower release schedule, say 
 3-4 major releases a year?


I would prefer to have an LTS release at most twice a year.

The first reason is that we get bugs like the installer not 
working properly. I reported this bug
https://issues.dlang.org/show_bug.cgi?id=18808
This is the sort of bug that enters when you release too 
frequently, and it's the sort of bug that causes potential users 
to give up and conclude the project has a quality problem.

The second reason is that the maintainers of libraries have to 
keep up with every release.

Reporting bugs is less likely if you need to install the latest 
release to confirm that it's still a bug. It's hard to keep up 
with language changes if you're a more casual D user like me.

Having LTS releases and making it easy for those wanting the 
latest and greatest to build the development versions would be my 
preference.
May 14 2018