Hello All,
Looking for resources on working on a very small footprint GNU/Linux for project we plan to start in the Affordable Solutions Lab. The idea is to put enough stuff on 32MB Flash RAM, in a thin machine that can run Mozilla and some apps, probably based on it using XUL.
We looked at uclinux, but that's not what we are looking for ... The project aims to squeeze a normal network enabled kernel, with display, VMM etc into as small a size as possible.
Any pointers?
Sameer.
On Thursday 31 July 2003 20:54, Sameer D. Sahasrabuddhe wrote:
Hello All,
Looking for resources on working on a very small footprint GNU/Linux for project we plan to start in the Affordable Solutions Lab. The
idea
is to put enough stuff on 32MB Flash RAM, in a thin machine that can run Mozilla and some apps, probably based on it using XUL.
We looked at uclinux, but that's not what we are looking for ... The project aims to squeeze a normal network enabled kernel, with
display,
VMM etc into as small a size as possible.
Any pointers?
VMM means i386? http://www.whitedwarflinux.org/
rgds jtdsouza@softhome.net
i think sameer's requirement is different.... he wanted to have GUI (i.e. xwindows , browser) in his system....whitedwarflinux cannot help.... it don't have xwindows in it...
sameer.... you can make or configure your own operating system with what ever features you want .... it is quite POSSIBLE..... you can make entire working system in less than 10MB of flash ROM.....
start with linuxfromscratch.org
regards
===== Ramesh, student, IIT, Bombay
__________________________________ Do you Yahoo!? Yahoo! SiteBuilder - Free, easy-to-use web site design software http://sitebuilder.yahoo.com
On Friday 01 August 2003 15:31, ramesh thatha wrote:
i think sameer's requirement is different.... he wanted to have GUI (i.e. xwindows , browser) in his system....whitedwarflinux cannot help.... it don't have xwindows in it...
Xwindows is gross overkill for such projects. Microwindows being developed for handheld and such others may be better. http://microwindows.censoft.com/
However one should have a very clear definition of the project goal else it will swiftly deteriorate into one more unusable tech toy.
rgds jtdsouza@softhome.net
rgds jtdsouza@softhome.net
On Sat, Aug 02, 2003 at 07:38:43PM +0530, J. T. D'souza wrote:
On Friday 01 August 2003 15:31, ramesh thatha wrote:
i think sameer's requirement is different.... he wanted to have GUI (i.e. xwindows , browser) in his system....whitedwarflinux cannot help.... it don't have xwindows in it...
Looked at whitedwarf ... its meant for the server, when what we are looking is meant for clients, or actually terminals with limited RAM, Flash, and no hard disk.
Xwindows is gross overkill for such projects. Microwindows being developed for handheld and such others may be better. http://microwindows.censoft.com/
This sure does look promising ... if it can really beat plain-vanilla X, then we are definitely interested.
However one should have a very clear definition of the project goal else it will swiftly deteriorate into one more unusable tech toy.
Well, I am not myself working on the project, so can't give too many details. But I do know its meant for the terminals I described above. Something like "application specific" small computers, which do only a handful of things, but do it well.
Some people in IIT were working on this as well ... they achieved to put a lot of stuff into 16MB of Flash. We hope to get inputs from them as well ...
Sameer.
On Saturday 02 August 2003 23:12, Sameer D. Sahasrabuddhe wrote:
On Sat, Aug 02, 2003 at 07:38:43PM +0530, J. T. D'souza wrote:
On Friday 01 August 2003 15:31, ramesh thatha wrote:
i think sameer's requirement is different.... he wanted to have GUI (i.e. xwindows , browser) in his system....whitedwarflinux cannot help.... it don't have xwindows in it...
Looked at whitedwarf ... its meant for the server, when what we are looking is meant for clients, or actually terminals with limited RAM, Flash, and no hard disk.
Xwindows is gross overkill for such projects. Microwindows being developed for handheld and such others may be better. http://microwindows.censoft.com/
This sure does look promising ... if it can really beat
plain-vanilla
X, then we are definitely interested.
However one should have a very clear definition of the project
goal
else it will swiftly deteriorate into one more unusable tech toy.
Well, I am not myself working on the project, so can't give too many details. But I do know its meant for the terminals I described
above.
Something like "application specific" small computers, which do only
a
handful of things, but do it well.
Some people in IIT were working on this as well ... they achieved to put a lot of stuff into 16MB of Flash. We hope to get inputs from
them
as well ...
A general purpose client ends up as a bad compromise between portability, usability, power consumption and costs. A case in point is the simputer. At a retail price of Rs. 15K it is most certainly too much for too little.
Devices for field personnel end up with data entry problems. Try using a PDA - with its nano sized, pico illuminated, virtually vanishing keyboard - in a crowded kalbadevi wholesale shop. Voice input devices require too much processing power. A digital voice recorder which can download the stored .ogg file into a backend server for voice recognition and processing is a far cheaper and better alternative. Infact u already have them - cellphones.
Designing a general purpose reconfigurable cell phone may be a much better idea. It will have a huge market and be commercially viable. Most of them have some avtar of arm processor, about 2MB of flash and 2MB of ram. And do nothing (apart from draining the battery) when you are not yakking. Java phones hitting the market are such devices. Only problem is that they run some jerks version of os and software, and requires third rate SDKs to reinvent perfectly usable existing applications. A linux cellphone will put real people in command.
rgds jtdsouza@softhome.net
On Sunday 03 August 2003 11:05, J. T. D'souza wrote:
On Saturday 02 August 2003 23:12, Sameer D. Sahasrabuddhe wrote:
If u intend to roll your own stuff use cml2 tool or the Qplus distro which uses cml2.
http://qplus.etri.re.kr/qplus-p/download.html
Peewee linux is also available but not much happening there.
i think we don't need any special tools to have our own embedded linux... I did from scratch... without xwindows (microwindows)..... i am able made it in 4.5 MB, this size is because i used glibc ..... if u use uclibc you can reduce size to 1 MB or even less...
in my system http server, telnet server are there... there is no hard disk in my system, i had only flash disk....
this embedded linux i designed for my project...
i think it is not a difficult task to include microwindows in that type system....
why i am stressing to come scratch is we will total control on the system (i.e we know everything about our system)
and one more thing to sameer....
when u r shifting from hard disk to flash disk.... you need to consider few things like.....
flash disk has limited number of write cycles... , but some daemons in linux box will continuously creates log files ... you cannot store those log files in flash disk... you have to redirect them to somewhere....
--- "J. T. D'souza" jtdsouza@SoftHome.net wrote:
On Sunday 03 August 2003 11:05, J. T. D'souza wrote:
On Saturday 02 August 2003 23:12, Sameer D.
Sahasrabuddhe wrote: If u intend to roll your own stuff use cml2 tool or the Qplus distro which uses cml2.
http://qplus.etri.re.kr/qplus-p/download.html
Peewee linux is also available but not much happening there.
__________________________________ Do you Yahoo!? Yahoo! SiteBuilder - Free, easy-to-use web site design software http://sitebuilder.yahoo.com
On Thursday 07 August 2003 10:28, ramesh thatha wrote:
i think we don't need any special tools to have our own embedded linux... I did from scratch... without
Tools are required for creating cross platform distros. Try doing this from scratch without tools. It is a pain in the ass and a waste of time.
why i am stressing to come scratch is we will total control on the system (i.e we know everything about our system).
CML2 lets you do precisely this.
and one more thing to sameer....
when u r shifting from hard disk to flash disk.... you need to consider few things like.....
flash disk has limited number of write cycles... ,
jffs2 file systems uses wear levelling and bad sector marking to ensure maximum usability. with current flash devices life is a couple of years - quite sufficient for most uses (unless u plan to go to Mars and comeback ;-).
rgds jtdsouza@softhome.net
D'souza,
thanks for your reply........and also special thanks to ur language.....But I had one fundamental doubt regarding last point.....
jffs2 file systems uses wear levelling and bad sector marking to ensure maximum usability. with current flash devices life is a couple of years - quite sufficient for most uses (unless u plan to go to Mars and comeback ;-). 99 percent industrial
applications has life cycle more than 2 years... I am not designing toys....
so u r saying that ...... life cycle is only 2 years...
two years what ever you r saying is very less.... suppose if am using flash disk in my industrial control system... I need to change it every two years...
Because of your design (according to u it is good ),,, suppose I am using around 50 industrial controllers at my work place..... I need to replace 50 flash disk every 2 years......
__________________________________ Do you Yahoo!? Yahoo! SiteBuilder - Free, easy-to-use web site design software http://sitebuilder.yahoo.com
On Sunday 10 August 2003 10:52, Ranganadh wrote:
Next GLUG Meet on 10th Aug. @ Ruparel College, Matunga Rd. (W), @
4.00pm
-------------------------------------------------------------------------------
D'souza,
thanks for your reply........and also special thanks to ur language.....
Err.. what is that ?
But I had one fundamental doubt regarding last point.....
jffs2 file systems uses wear levelling and bad sector marking to
99 percent industrial applications has life cycle more than 2 years... I am not designing toys...
In which case using X86 based designs is not a very good idea.
so u r saying that ...... life cycle is only 2 years...
By couple of yrs I mean 5 ~ 7 yrs. It depends on number of writes to the flash. A rough example if u r using a 20Ksps 12 bit ADC your data rate is 40KB/s. with a maximum erase write cycle of 1000000 and wear levelling, device life will be 4.75 years. If u use JFFS2 which has built in compress/decompress life will be higher. On the other hand u will have many more ADCs and sensors, hence higher write loads so reduced life..... etc. etc.
two years what ever you r saying is very less.... suppose if am using flash disk in my industrial control system... I need to change it every two years...
That is nonetheless better than harddisks in an industrial environment.
Because of your design (according to u it is good ),,,
A design is only as good as the underlying assumptions made about it's environment and use.
suppose I am using around 50 industrial controllers at my work place..... I need to replace 50 flash disk every 2 years......
More like 4~6yrs - quite manageable. The real problem is in unmanned remote installations.
rgds jtdsouza@softhome.net
-
J. T. D'souza -
ramesh thatha -
Ranganadh
-
Sameer D. Sahasrabuddhe