On Jan 22, 2010, at 9:31 AM, Dima Pasechnik wrote:
Robert,
the advantage is that it will simplify the *development* of Sage.
Ironically, I personally find shipping all our dependancies makes
development easier--I don't have to worry about someone else using
different version than I have, and it's easy to look at and, if need
be, patch the source directly.
Right now lots of stoppers seem to come from upstream packages.
For porting, yes. Whether or not depending on people to build their
own upstream packages, then dealing with the issuing dependancy issues
is a gain is up for debate. Probably depends on the package.
I also do not see a real problem with "specific versions" of
packages. Somehow,
all the other open-source math projects seem to be able to manage this
well,
e.g. Singular manages to coordinate with GMP.
As well, lots of things like needlessly tying Sage up to a very
particular version or
an environment can be sorted out simply by using autoconf properly...
Some packages, like GMP, have a well defined and strongly backwards
compatible API. Other packages, like GAP and maxima, aren't even
considered libraries, and freely make changes that are semantically
inconsequential but change how they interact with the "user" (which in
this case is a pexpect interface, not a human who won't notice when a
question is re-phrased). Even when changing versions doesn't give
errors or bad results, often doctests will fail. (For example, the
Debian port fails doctests all over the place, and trac is full of
examples where bumping to a newer upstream version involves non-
trivial work.) Still other packages are arcane, specialized sets of C
files that where the author has little or no interest (or ability) in
getting into or maintaining as part of a mainstream repository archive
(if it would even be accepted).
I'd love for someone to be able to do apt-get sage, and get a
reasonably modern, fully working version (without too much redundancy
with what's on their system, especially if that helps us get into
debian, etc.) As a concrete goal towards that end, I'd suggest making
a stripped version of sage that omits "easy" packages such as bzip2,
zlib, maybe even gmp/mpir that still passes all doctests, modifying
the prereq script appropriately.
- Robert
--
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org
