![[info]](http://l-stat.livejournal.com/img/userinfo.gif){kind=small}
Targets
Really good mail from Bill Nottingham about target markets for Fedora. It is pretty crazy how much software in general is written without thinking about who will be using it, why they'll want it, and how it will be used.
A current example of the wrong way (in my opinion) to create software is OpenSolaris' NetworkManager-type program. They call all this work design, but really almost all of it is the engineering implementation design, especially this thread. It has very little to do with how people use the software; there are some sentences to the effect of "oh yeah, there will be GUI popups or something". The UI page doesn't exist yet even. To be fair to them, maybe I'm looking in the wrong place, but I didn't see any non-engineering design.
Having been next to Bryan, Seth and Dan when they were working on NetworkManager, I saw firsthand how a good interaction design and an understanding of the target market is key. For example, in the initial design, they made a decision not to support static IP addresses. That may seem crazy to a programmer, but the target market (Fedora wireless users) basically never uses static addresses. Thus, no need to support them, and the design became much simpler.
Nowadays, I wouldn't even try to write real software unless I had a good interaction designer on the team. From the start. Not like someone pulled in 70% of the way through to double check the widget spacing on all the incomprehensible popup dialogs I'd created.
A current example of the wrong way (in my opinion) to create software is OpenSolaris' NetworkManager-type program. They call all this work design, but really almost all of it is the engineering implementation design, especially this thread. It has very little to do with how people use the software; there are some sentences to the effect of "oh yeah, there will be GUI popups or something". The UI page doesn't exist yet even. To be fair to them, maybe I'm looking in the wrong place, but I didn't see any non-engineering design.
Having been next to Bryan, Seth and Dan when they were working on NetworkManager, I saw firsthand how a good interaction design and an understanding of the target market is key. For example, in the initial design, they made a decision not to support static IP addresses. That may seem crazy to a programmer, but the target market (Fedora wireless users) basically never uses static addresses. Thus, no need to support them, and the design became much simpler.
Nowadays, I wouldn't even try to write real software unless I had a good interaction designer on the team. From the start. Not like someone pulled in 70% of the way through to double check the widget spacing on all the incomprehensible popup dialogs I'd created.
(Anonymous)
Agh that is stupid
NIH syndrome? Those guys should really know better in the first place.
(Anonymous)
Re: Agh that is stupid
(Anonymous)
have a look at the actual project page...
http://opensolaris.org/os/project/nwam/
nwam is going to be the way networking is configured on Solaris - even if you're running without X, or on CDE. Its a superset of what NetworkManager provides. GUI is only part of it.
Have a look here for info on the UI design:
http://opensolaris.org/os/project/nwam/U
The mail thread you linked to is an architecture review (i.e. engineering implementation) - its not a UI design review.
-Mark
(Anonymous)
Re: have a look at the actual project page...
This might be similar to what you meant by target markets:
http://opensolaris.org/os/project/nwam/s
-Mark
Re: have a look at the actual project page...
Now it just seems likely to me to be doomed by their attempt to support every possible network configuration.
(Anonymous)
Re: have a look at the actual project page...
Re: have a look at the actual project page...
If I was a server administrator, I'd want a nice set of shell tools, not that.
Personally, I think the right approach in the short term is to:
1) Ship NetworkManager on desktops
2) Use whatever the OS currently provides for servers
There is certainly some potential for unification of some system components, but honestly, some of the requirements listed by some of the kernel engineers are just wacky. The second and third of Darren's (why not just give him a UI where he can craft his own DHCP packets in a hex editor?), for example. And Dan's "I run my own enterprise network at home".
Anyways it's not my place to debug OpenSolaris process, I'll let it be =)
(Anonymous)
Re: have a look at the actual project page...
As regards 1): It might even be possible to "port" NetworkManager to use nwam
As for 2) nwam will become what Solaris uses for network configuration and there will be a GUI for that. Most users won't have to muck around in configuration as nwam will mostly just "Do The Right Thing" (and provide feedback to the user). For those admins who want more control they will have the option of a GUI.
-Mark
(Anonymous)
Re: have a look at the actual project page...
For servers perhaps.. For ordinary users it can not be that simple as 2/3 of the world's sold computers are laptops and that means often, very often wifi.. With certificate selection (802.11x), encryption types, selecting out of multiple possible networks (not all might be correct or work, some internal use only etc), querying for passwords, ...
Good luck with that, get back to me at 2050 or so when you got the first actually working version of your idea, that actually serves any users well, completed.