2008-01-14 14:07:12 ~~Many Thanks~~
Kris Dickie (CANADA)
Message: 49562 I think it's important to give the BF-uClinux support team credit. I run a support forum for my company as well (although much less traffic), and I know how demanding it can become when developing and giving support are both full-time tasks. Without their help, I don't think any of the Linux based BF projects would go anywhere. The ever-evolving Linux kernel is not an easy thing to master, and any steps we as users make, are great accomplishments as well.
Thanks for your continued support, look forward to seeing this project on distrowatch someday
Also, as a note to all the BF-uClinux users out there: if you switch to a Linux operating system for your desktop (or at least run a dual boot), and use it daily, your progress on understanding embedded Linux projects will increase ten-fold. There is no excuse not to run Linux these days, when it is so easy to install, and comes bundled with so many amazing free and open-source applications; many of which, run circles around the commercial and MS based applications.
2008-01-14 21:07:06 Re: ~~Many Thanks~~
hua zhu (CHINA)
Message: 49569 my thanks follows...
they're really doing a great work.
2008-01-15 09:47:30 Re: ~~Many Thanks~~
Graham Davies (UNITED STATES)
Message: 49599 Kris Dickie wrote: "... if you switch to a Linux operating system for your desktop (or at least run a dual boot), and use it daily, your progress on understanding embedded Linux projects will increase ten-fold. There is no excuse not to run Linux these days ...".
I don't want to start an argument or anything, but I don't really agree with this. By all means, do run a Linux box if you can. There is one remaining excuse not to, which is "My IT department won't let me", but that aside, it really is pretty easy. Try xubuntu for a light-weight distro. But ... I have been using UNIX since 1992 and Linux since 1995. I have built a Linux distro from scratch following <a title="Gerard Beekmans" href="http://en.wikipedia.org/wiki/Gerard_Beekmans">Gerard Beekmans' book. I'm still befuddled by uClinux. I think that everything you can learn from running Linux on the desktop is basically a pre-requisite to using embedded Linux effectively. Linux neophytes shouldn't even try. But, the needs of an embedded system with limited memory and critical real-time constraints are completely different from the needs of a desktop workstation with a 4 GHz processor, 2 Gbytes of RAM and nothing to do but recalculate a spreadsheet and move words around in a word processor.
Oh, and add my thanks to the support team.
2008-01-15 12:04:46 Re: ~~Many Thanks~~
Robin Getz (UNITED STATES)
Message: 49603 Graham:
> don't want to start an argument or anything, but I don't really agree with this.
I assume that the only part you don't agree with is:
> There is no excuse not to run Linux these days
> if you switch to a Linux operating system for your desktop, and use it
> your progress on understanding embedded Linux projects will increase ten-fold.
What I have found is that people that are more comfortable with Linux in any capacity, have an easier time with Linux in an embedded target - whether it be with uClinux - or embedded Debian. I can't quantify ten-fold, but I can say the questions we get from people comfortable with Linux on a host, are at a different level than people who are not. (mainly because the people who run Linux on desktop, can try out what they are trying to do on a Blackfin on their host, and if it doesn't work there - it will not work on the Blackfin either - so self help answers more questions).
You also said:
> I'm still befuddled by uClinux.
Which part is befuddling? If there is a section of documentation that is missing, or if we are not doing things clearly, I would be happy to try to put something together to explain things.
2008-01-15 13:20:14 Re: ~~Many Thanks~~
Graham Davies (UNITED STATES)
Message: 49604 I only disagree with the part of the original message that I quoted in my reply. I think if you re-read what I wrote it should be pretty clear what I'm trying to say. Please don't reply as if my entire post was "I don't really agree with this". I am very grateful to the support team and I agree completely with the OP that being responsible for support and doing development at the same time is very difficult and those who do it well are most admirable people.
The paragraph "What I have found ..." in your reply seems to agree more or less with what I said, particularly the part about questions being on a different level. As I said, I don't want an argument, so I'm not going to repeat what I wrote before for fear of starting one.
There are many parts of uClinux that befuddle me, it won't help to single anything out. The section of the documentation that is missing is an exhaustive User Manual that identifies and explains every last feature, tells me whether I need it and, if so, how to use it. I'm trying to find a simple RTOS in a gigabyte of almost completely undocumented source code. I'm not convinced I even need pre-emption for this job. It's like I have a letter to put in a post-box a mile down the road and I'm told I have to learn to fly a helicopter there to do it.
All I'm saying (oh, God, he is going to repeat himself after all) is that 15 years with a UNIX box on my desktop has not prepared me for the deployment of uClinux in a real-time embedded system. (By the way, I have used pSOS, vxWorks, uC/OS and have written an RTOS from scratch.)
2008-01-15 14:18:30 Re: ~~Many Thanks~~
Robin Getz (UNITED STATES)
Message: 49607 Graham:
I'm just trying to understand where we need to improve - not start a battle - so no need to be defensive - I'll be the first to admit that sometimes we suck. sometimes the documentation sucks, sometimes the source code sucks - most times not - sometimes yes - I'm just try to make it better. (Thanks are nice - and very appreciated, but I would rather understand where we suck - so we can improve, and suck less).
>The section of the documentation that is missing is an exhaustive User Manual that identifies
>and explains every last feature, tells me whether I need it and, if so, how to use it
I'm not sure I can make exhaustive... Linux is a pretty big topic to write "the only Linux manual that you will ever need"
I try not to replicate chapters of already existing books in our documentation, and these are the ones I have on my shelf (my self is pretty big).
The docs site trys to fill in the rest - but even with 16,811 books that come back on Amazon when you search for the topic "Linux", there are still quite a few gaps that people run into. (Plus, as soon as the book is published, it is out of date).
If you need a one day getting started tutorial - ask your local ADI contact - they should be able to help with the basics. There are a good detailed training classes (4 days- http://www.sysdcs.com/ ) that should cover more detailed aspects as well.
It is hard to document everything - for example busybox - is a project by itself - with it's own documentation. We don't try to replicate what they have created - it's duplication. but I can understand that this causes some confusion - going to 100 different documentation projects to get a project out the door...
I'm not sure how to solve that one - it is a difficult problem.
2008-01-16 10:37:19 Re: ~~Many Thanks~~
Graham Davies (UNITED STATES)
Message: 49671 Gee, I don't think I'm the one being defensive here, but let my give my answer to your question.
First, you do not suck, ever. This whole thread was about how great your work is and how grateful everyone is (or at least should be) to you. If anyone is unhappy with your answers, they should spend a day getting interrupted with random questions every 20 minutes with no reduction in their workload and see how much patience they retain.
So, where to improve? Well, Linux is what Linux is and you're not going to fix scattered documentation and obscure and fragile source code in a 1 Gbyte Linux distribution. Nobody should be here if they don't understand what Linux is (although, of course, we are, because our managers make this decision, not us, and they hear "free" not "fragile"). The guys who need Linux for the applications (networking, VoIP, sound, etc.) are probably in good shape. They're basically building tiny Linux boxes with few constraints over a desktop machine, other than size and power which is where the Blackfin comes in. In my opinion, though, the guys who are here because nobody wants to pay for VxWorks or uC/OS when you can get an embedded operating system for free are not in such good shape. Me, for example. In terms of function, uClinux provides about 10% of what I need, with the rest coming from code I write myself. In terms of size, however, uClinux takes up more than 90% of my binary image. My source files (which, by the way, are more than 50% comments) total less than a megabyte, which is 0.1% of the distribution. If anyone out there has worked with CVS, they may have some appreciation for what it's like to manage a code base like this. My pathetic little concrete suggestion is this - offer a second distribution of uClinux for the Blackfin that is cut back to the bare minimum from which people can build *up* to the function they need. If you have time to add more documentation, add it about this. Or, let everyone participate by putting up a Wiki (with administractive approval of editors).