Re: [Chicken-users] Subversion support misguided IMHO

[Brandon J. Van Every]

Peter Bex wrote:

On Fri, Jan 05, 2007 at 01:07:31PM -0800, Brandon J. Van Every wrote:
  

If I were making the decisions on such things, I would drop Subversion 
support and force people to use / learn Darcs.  I'd provide a HOWTO, of 
course.
    

I'm not using Darcs this lifetime, at least as long as it has a ghc
dependency.  If you use a slightly odd platform (in my case, NetBSD/macppc
and NetBSD/amd64) where the ghc guys don't build binaries for, they require
you to get ghc sources, a ghc binary for the same architecture but another
OS, build intermediate C files specifically for that architecture
(yes, their generated C code is architecture-dependent!), move the source
tree to the target OS and compile it.
  

Hm, I agree that not being readily available on all architectures is a liability, for a language project like Chicken Scheme.  It wouldn't be for a more mainstream application, which could be expected to reside on more mainstream OS configurations, but languages need to reach far and wide.

This also means it can't be easily packaged with automated packaging systems,
and I have to do this every time I need to update ghc to a new version for
any architecture they don't care to provide binaries for.

I don't like having to jump through hoops, especially not to build a
frickin' compiler just to get some source control program on my system.
  

Well, I have no inherent sympathy as far as making people build from source.  That's all that Chicken currently provides, after all, and we've previously discussed how burdensome it would be to provide a generalized binary capability.  The difference is that the Chicken builds *actually work*.  Apparently, GHC builds are dicey.  That's no fun at all.  That's why I spent 1 year making Chicken's build "not suck."  The vast majority of builds out there, totally suck.

(besides, I don't even have another OS on both computers, nor access to
 one that does)

That's an issue that gets solved when one has an actual need.  I've had HD space for an additional 2 OSes for a long time.  But I certainly didn't need additional support burdens and confusion.  Even now, poking at IBM Cell programming as I am, it's going to be awhile before I see Linux as my primary development environment.

  I'll just use svn (or any one of the many other source
control systems out there, distributed or not), thanks very much.
  

The issue to me is not so much the inherent merits of Darcs vs. Subversion.  I'm inclined to believe that Darcs is better for distributed open source development than Subversion, as it has a more exotic patching capability, but I haven't closely investigated Subversion.

The issue to me is when the development community is split into different toolchains.  If I were running a project, I would declare that unacceptable and make everyone use 1 thing.  I think it is important to socially engineer people, so that they know the tools, and do not have futzy barriers in the way of making contributions.  Also so that people have shared expertise, rather than failing to help each other because 3 people know Darcs and the rest don't.

I am open to ditching Darcs, if there's something as technically advanced, and mature, that has a greater platform reach, that we can all agree on.  If that also allows us to ditch Subversion, so that everyone is on the same page.


Cheers,
Brandon Van Every