Hi Bogdan,
<Oct-Dev ML added in TO: list. Pls do "reply all" when answering>
Bogdan wrote:
> Hi,
>
> There is an alternative, of course. The MinGW folks have had an
> installer for some time, mingw-get (and a graphical version of it, which
> is WIP). This can install, handle dependencies, uninstall, update, etc.
> packages; everything from MSYS to GCC and bzip2 is distributed in this
> fashion---which is part of the reason why it may seem that there aren't
> too many MinGW installer downloads. Basically, it can do everything that
> is expected of a package manager.
There's a precedent: Cygwin.
As far as Octave-on-Windows goes, Cygwin has never been as successful as
the complete "archive installers" and NSIS installers.
Why would MinGW-get perform better?
> How about providing such a package for Octave? The package consists of
> an XML catalogue and one or more archives for different components of
> the package (e.g., one may wish to separate documentation, binaries,
> license stuff, extra stuff, etc.). This seems like the perfect solution.
Of course the usual insipid answer: It's up to you. This is a volunteer
project, remember. Please go ahead.
To have a start:
All dependencies for core Octave have been built by Tatsuro Matsuoka
based on a.o., gcc4.6.2 and can be found here, together with build
instructions:
http://www.tatsuromatsuoka.com/octave/Eng/Win/
Then, Nitzan compiled and built later Octave version binaries, + built
almost all Octave Forge packages. AFAIK he patched & amended a few
things but unfortunately he never clearly outlined those fixes.
(That is, Nitzan and I use Tatsuro's build stuff but N. gets better test
results. And he can build much more OF packages than me.)
<SKEPTIC MODE on>
But even if you proceed this way (which we all would applaud I think),
the points in my previous e-mail still apply:
1. Focus
(Barring download limits on SourceForge) Please consider balancing
precious developer time for core Octave and Octave-Forge development,
versus investing precious developer time for building a package mgmt
system from scratch and maintaining it.
The weak point here has always been lack of developers for Windows. We
are very lucky at the moment to have about 2-3 names "active", incl. the
one building the MSVC binaries.
2. User perception
(The main point with Octave on Windows) Rare exceptions aside, the
intended audience isn't known for patience with and insight in pkg
management systems (see Cygwin). All "they" know about is "one-click
installers". And why not? Choice enough: Freemat, Scilab, ML itself, etc.
This audience has simply made its choice already a long time ago: easier
& less dependence = better. Good luck educating them otherwise.
3. Intended gain
(Referring to your 1st post about unneeded code) What's the point of
having a package mgmt system, if for the vast majority of intended users
-who start from scratch- it makes almost no difference whether they get
about the same files and same amount of bytes via a package manager or
from a monolithic installer?
Some simple weeding in the current .7z archives may be much more effective.
Another thing here: We regularly get reports that Octave is used in
courses. Imagine tens or hundreds, or more, students all trying to
install Octave simultaneously over the school/uni network using a
package manager d/l-ing from SF. That school/uni had better have a
monolithic Octave installer somewhere on the LAN, and that would be
better for SF's bandwidth, too.
To summarize:
What do you want to achieve with trying to improve Windows-Octave's
distribution/installation mechanism?
</SKEPTICAL MODE>
Philip
------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Octave-dev mailing list
Octave-dev@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/octave-dev
