Tanglin With Alpha

[David Hutto]

Just as a quick survey...Are individual programmers happier with tangling with the alpha code, or more comfortable with beta versions, or does it matter, dependant on the circumstances?

 

[Chris Angelico]

Just as a quick survey...Are individual programmers happier with tangling with the alpha code, or more comfortable with beta versions, or does it matter, dependant on the circumstances?

Depends on the program, enormously. Or, to put it another way, depends
on whether I'm trying to "just get the job done" or am prepared to put
in some effort.

I run a web and mail and etcetera server. For that, I use Apache,
bind9, courier-imap, and a bunch of other programs, all obtained
through apt-get.

I also am the author of a MUD client, written in Pike. Because of the
nature of what I'm doing there, I tend to push the boundaries of the
language itself; that means I sometimes find bugs, submit patches, and
all that. So currently, I'm running a bleeding-edge Pike that consists
of the latest from upstream plus one patch of my own that hasn't yet
been accepted... so it's "from __future__ import socket_nodelay" if
you like. The price I pay for that is that, sometimes, stuff's broken.
I try to import the bittorrent client and boom, it fails because
something's half way through being edited in the SSL code and it
wouldn't load. That's something I choose to accept with Pike, but I
would *not* accept it with, say, gcc. With the C compiler, I expect it
to just work.

But alpha and beta versions? Almost never. I'm currently running
Python 3.4.0b2 on Windows, because I don't have facilities to build
Python from source on Windows; on my Linux boxes, I use Python 3.x
straight from Mercurial (same as with Pike above), or else whatever I
can apt-get. Either I'm prepared to use a development version or I'm
not, with very VERY few exceptions (I think I once built a Linux
kernel from a development tarball - really no point, I could have done
just as well going from git).

Ultimately, it comes down to how much work you want to do versus how
much trust you want to place in someone else. If you'd rather trust
someone else, take a published stable version (especially one that you
can get from a stable OS distribution's repository - not only is it
convenient to apt-get everything, you can be confident that the Debian
Wheezy repo has stuff that's known to work with Debian Wheezy); if
you'd rather do the work yourself, build from source, at whatever
point you like.

ChrisA