Main Area and Open Discussion > General Software Discussion

mswin vs linux in academia

<< < (3/3)

jgpaiva:
Here's my opinion as a phd student in the systems area:
I believe it has absolutely no relation with ease of use of the operating system. As 40hz mentioned, at this level everyone could learn any of the OSs with ease.
For the guys around me, it seems to be a combination of several factors:
1 - there's a significant larger body of work for unix/linux (more people are using it, which makes others change too). example: my experience is that the software available from recent years of the Symposium on Operating Systems Principles (the top conference in the Systems area) is mostly prepared to be ran in linux (despite much of it being cross-platform).
2 - we do much stuff that builds on industrial work (I'm thinking of Apache Cassandraw or red hat infinispan) for which it is easier to find support for the linux platform.
3 - we build some stuff for the industry, which prefer to receive stuff for their platform, typically linux.
4 - our processing grids are in linux (using Condor High-Throughput Computing Systemw). I don't even know if there's anything comparable in windows.
5 - I have this notion that for those who work lower on the stack (kernel level?), it's either harder or impossible to do it in Windows due to being a closed platform (I have no experience in this and may be wrong, though).
6 - much of our experience during the university is directed at linux/unix.
7 - the availability of the tools we use (gnuplotw and graphvizw are the only ones that occur, but there are many others): almost everything is an apt-get away from being installed, and if necessary we can install them in the servers to run with larger datasets. In windows, they are either harder or even impossible to install.

For the reasons above, frequently people use either linux or mac for their desktops, and linux for the servers. It is also frequent to go to meetings with other universities and find the whole room filed with macs. Last year's SOSP was here in Lisbon and I attended it, and my experience was that in general mostly only the people working at MS research use Windows and the percentage of people using mac is much larger than in the general population.

Having some kind of unix on the desktop simplifies the interactions with the servers (in our case, we have about 28linux and 1dual-boot (linux+windows), without taking into account the computing grids which are all linux). I have in the past used windows with cygwin to achieve a similar result, but felt that I was trying to "patch" windows making it look more like linux instead of using the real thing, and eventually had to change.

There's a notable exception to everything I said above: there are a few guys in our research group who had a few microsoft scholarships some years ago. Those guys have built stuff for windows for those projects and are still using it today as their main desktop.
Also, this applies only to the Systems area. There are other areas which I suppose are not as tied-in to linux since their tools are more windows-centric. (I'm thinking of the people-machine interface guys or information management. But I may be wrong :P )

Tuxman:
Virtually all GNU software - and even some BSD software - has been ported to Windows. There is no reason to prefer Linux just for its applications.

mahesh2k:
I meant that exclusively in relation to creating software distribution packages, i.e. installers, setup files, packages, whatever anyone wants to call them. If you know something I don't about how to sanely create a package for Linux, please, do tell me~! :)
--- End quote ---

How build system has to do with simplicity of the OS? Software on linux is routed via repository and software center these days, installers are hardly needed in these cases for popular programs as community takes care of it. And this method is much easier and better than windows software installation. Hell, I am sure they'll start with repository concept soon. Again, You're comparing build package system for multi-distro linux to that of single standing windows. I am sure you do know that there are formats specific to the distribution and then there is build from source method. Software like gDebi helps in any case and user has to do nothing other than downloading the archive. To get to your point about better packaging system, there was Installjammer, for both windows and linux(or *nix) which is now discontinued and then there are some commercial installers and free like installanywhere and autopackage.

Edvard:
...
If you know something I don't about how to sanely create a package for Linux, please, do tell me~! :)
-Renegade (August 02, 2012, 10:17 PM)
--- End quote ---

Mahesh2k is right, most distros use repository + package management for installing and the like, so you'll have to get comfy with the workings of Apt and Rpm to target the majority of those systems.

The only other ways I've seen:
-Plain compressed package (tar + gz or bzip2) that unpacks to the most common default locations for executables and config files.
-Shell script that contains a binary payload, which is probably the closest thing Linux has to Windows stand-alone installers.
-0install. Never did figure that one out...

That said, I've gotten pretty cozy with the Apt package management system, and made a few packages that actually worked, so maybe I can give a few pointers.  It's not that hard, just the documentation makes it seem like it.

As far as Linux in academia, jgpaiva has the best comments on that; Linux was basically born out of Unix in an academic setting, so it doesn't surprise me at all that the tradition continues with Linux.
I would expect more Windows usage in a setting where the focus was development for the platform.

Navigation

[0] Message Index

[*] Previous page