[9fans] SCMs

[9fans] SCMs

[Lyndon Nerenberg]

>> I struggle to understand how version control is not more actively used.

> It's not particularly necessary when you have global state with
> snapshots provided by a shared WORM fs.

I always thought that argument was a bit suspect.  And with the loss of
sources.bell-labs.com, it's apparent why.  The only revision history was
in the venti.  Now that that is lost, so is that history.  I know that
there are partial mirrors of sources, but none go all the way back to the
dawn of the sources venti archive.  And on the mirrors, we lose the
'blame' functionality fossil provided by tracking who last updated a file.

If this had been hosted in an SCM, it would have been so simple to
replicate that full history elsewhere.

The other bit that snapshots/dumps miss is context.  When everyone working
on the code was within shouting distance of the "unix room" that wasn't an
issue.  But now, that context has been lost.  Annotations about the "why"
of a commit are as important as the "what."  diffy(1) answers the "what,"
but not the "why."

> DVCS adds a lot of complexity
> for questionable gain, in that environment.  9front's adoption of
> mercurial is a historical accident rather than a desired outcome.  But,
> I understand that most people just want to use the tools they already
> know.  It's much easier than learning a new paradigm.

+100 on DVCS and needless complexity.  cvs or sccs provides all the
functionality I've ever needed in an SCM system.  Although I confess I
have been seduced by git's ability to instantly create and switch between
branches.  It makes trying out "what if" scenarios completely painless.
But it's not enough to convince me to use git except on very rare
occasions.

--lyndon

Re: [9fans] SCMs

[Erik Quanstrom]

[-- Attachment #1: Type: text/html, Size: 252 bytes --]

Re: [9fans] SCMs

[Bakul Shah]

On Tue, 13 Feb 2018 15:13:36 -0800 Lyndon Nerenberg <lyndon@orthanc.ca> wrote:
>
> +100 on DVCS and needless complexity.  cvs or sccs provides all the
> functionality I've ever needed in an SCM system.  Although I confess I
> have been seduced by git's ability to instantly create and switch between
> branches.  It makes trying out "what if" scenarios completely painless.
> But it's not enough to convince me to use git except on very rare
> occasions.

Git has far too many subcommands and impementation warts but
its underlying model has just 4 types of objects. The
distributed part of git is conceptually fairly simple too.
Over time I have used sccs, rcs, cvs, subversion, mercurial,
git and a couple other SCMs. Each has its pros and cons.  By
now I primarily use git because I can run it on all the
platform and it is quite fast. And there is an active
ecosystem around it. Things like code review systems do help
quite a bit with improving code quality (in my last job I used
gerrit to brainwash people into following good practices by
providing detailed core review comments).

Dave MacFarlane's git client (dgit) does a decent job on plan9.

Re: [9fans] SCMs

[Rui Carmo]

> On 14 Feb 2018, at 01:47, Bakul Shah <bakul@bitblocks.com> wrote:
> 
> Dave MacFarlane's git client (dgit) does a decent job on plan9.

This interests me greatly. Last time I checked there wasn’t a good enough got client, so I used Mercurial. Where is dgit exactly?

R.

Re: [9fans] SCMs

[Dave MacFarlane]

[-- Attachment #1: Type: text/plain, Size: 604 bytes --]

On Feb 14, 2018 02:22, "Rui Carmo" <rui.carmo@gmail.com> wrote:



> On 14 Feb 2018, at 01:47, Bakul Shah <bakul@bitblocks.com> wrote:
>
> Dave MacFarlane's git client (dgit) does a decent job on plan9.

This interests me greatly. Last time I checked there wasn’t a good enough
got client, so I used Mercurial. Where is dgit exactly?

R.


https://www.github.com/driusan/dgit/

It's written in Go, which means it'll only work on platforms that Go
supports (I think there's a list somewhere on the Go wiki, but it's some
subset of 386/arm/amd64 depending on which fork you're using.)

[-- Attachment #2: Type: text/html, Size: 1300 bytes --]

Re: [9fans] SCMs

[Lucio De Re]

On 2/14/18, Dave MacFarlane <driusan@gmail.com> wrote:
>
> https://www.github.com/driusan/dgit/
>
> It's written in Go, which means it'll only work on platforms that Go
> supports (I think there's a list somewhere on the Go wiki, but it's some
> subset of 386/arm/amd64 depending on which fork you're using.)
>
Let's not forget  Cherry's plan9/mips64 port of Go. I seldom test it,
regrettably, but I've yet to be disappointed. I own a Lemote Yeeloong
and I'd use it more if I had a clue on how to get a disk driver
working on it. But it's OK net-wired to a file/cpu server.

Cherry's amazing development skills unfortunately seem totally consumed by Go.

Lucio.

Re: [9fans] SCMs

[Steve Simon]

fyi

i spoke too soon, the labs website went a while ago, but the sources machine has returned, well i was able to access it last week.

last chance (i suspect) for those wanting to download the contrib dirs before they disappear - i got mine 😀

-Steve
 

On 13 Feb 2018, at 23:13, Lyndon Nerenberg <lyndon@orthanc.ca> wrote:

>>> I struggle to understand how version control is not more actively used.
> 
>> It's not particularly necessary when you have global state with
>> snapshots provided by a shared WORM fs.
> 
> I always thought that argument was a bit suspect.  And with the loss of sources.bell-labs.com, it's apparent why.  The only revision history was in the venti.  Now that that is lost, so is that history.  I know that there are partial mirrors of sources, but none go all the way back to the dawn of the sources venti archive.  And on the mirrors, we lose the 'blame' functionality fossil provided by tracking who last updated a file.
> 
> If this had been hosted in an SCM, it would have been so simple to replicate that full history elsewhere.
> 
> The other bit that snapshots/dumps miss is context.  When everyone working on the code was within shouting distance of the "unix room" that wasn't an issue.  But now, that context has been lost.  Annotations about the "why" of a commit are as important as the "what."  diffy(1) answers the "what," but not the "why."
> 
>> DVCS adds a lot of complexity
>> for questionable gain, in that environment.  9front's adoption of
>> mercurial is a historical accident rather than a desired outcome.  But,
>> I understand that most people just want to use the tools they already
>> know.  It's much easier than learning a new paradigm.
> 
> +100 on DVCS and needless complexity.  cvs or sccs provides all the functionality I've ever needed in an SCM system.  Although I confess I have been seduced by git's ability to instantly create and switch between branches.  It makes trying out "what if" scenarios completely painless. But it's not enough to convince me to use git except on very rare occasions.
> 
> --lyndon
>