aufs + opkg

[johnbock]

Attached are two, for what I believe, very import packages for freetz

1. aufs - another unionfs: aufs is safer, smaller and faster than unionfs
2. opkg - packaging system: ipkg is not being actively developed

The corresponding packages contain

• patches for freetz configuration
• freetz build directories
• tar.bz2 for the dl directories

Both of which have been minimally tested. I hope to see both of these in freetz-1.1

[Silent-Tears]

The filesystem seems to be ok, but i do not see any reason to implement a package format like opkg, cause we do not support dynamically loaded packages at the moment.

[johnbock]

but i do not see any reason to implement a package format like opkg, cause we do not support dynamically loaded packages at the moment.

I agree that freetz does not yet support dynamic packages. That's one reason I wanted it in freetz-1.1; if you like we could put it in freetz-future. As far as I've understood the latter should be some kind of sandbox.

In this way we could specify and prototype the build system.

[Silent-Tears]

I think dynamically loaded packages not for freetz. Danisahne said some time ago (month) that he works on somethink like that. The progress you can see on his sourceforge-page.
How do you think we should use dynamically installing? The Flash is RO at runtime, is it?

[dileks]

Attached are two, for what I believe, very import packages for freetz
[list=number][*]aufs - another unionfs: aufs is safer, smaller and faster than unionfs[*]opkg - packaging system: ipkg is not being actively developed
[...]
Both of which have been minimally tested. I hope to see both of these in freetz-1.1

After months of being quite about me, I must(!) manifest my point-of-view.

First, to the person johnbock.
A highlight to have win you for freetz.
The first amazing topic was to initiate the Shell-Coding-Style-Guide.
I can't find it on the freetz-Wiki, but I know I boomarked it.
There comes a lot of good ideas and solutions from your corner.

To the topic AUFS and OPKG:
I hear the first time of "opkg", but I already discussed ipkg with the main developers.
It is a fact that it is quite around danisahne, the man made his diploma and has its business starting in a "real" (paid-for) job.
That's why freetz was born (remember that).
To AUFS I can say, I applied/used it in my own Linux-LiveCD "SEL" based on Sidux/FLL with e17 as desktop-(window)-manager.
"SEL" has the ability of been created with daily dist-upgrades.
AUFS is an overlay system where you not only "pretend" a writable filesystem - you can!
I didn't look in your attached archives: Did you backported a newer issue of AUFS or a suitable one for kernel v2.6.13.1?

In conclusion:
The idea of dynamic packaging is more than overdue.
If the base-system is under 4MBytes (fits in FLASH-RAM) we have solved another big problem on elder TI-ar7-based routers!
Currently, I am involved in other projects.
But, I have an eye on it.

To the developers:
What speaks against to put and test it in a different branch (freetz-future)?

Kind Regards,
Sedat

[johnbock]

How do you think we should use dynamically installing? The Flash is RO at runtime, is it?

To say it simply, depends on the topology that's being used. For quite sometime now I've been using a mixture of flash in squash and an overlay. My overlay is at this time mini_fo on tmpfs.

Although, this has no advantages in a production system, I've been using it to test stability and to prototype an usbroot/mini_fo(aufs) for the 7270. Which of course allows a much more flexible flash/overlay scheme and dynamic packaging...

[johnbock]

Did you backported a newer issue of AUFS or a suitable one for kernel v2.6.13.1?

Unfortunately, not. I have only a 2.6.19 based box: 7270...

[olistudent]

Hi.
Can you point out the advantages from aufs over mini_fo?

Greets Oliver

[johnbock]

Can you point out the advantages from aufs over mini_fo?

To be honest, I don't know enough of mini_fo to list the technical advantages of aufs. However, the greatest plus that I see is that it is being actively maintained by a large and growing community.

[McNetic]

Yes, Danisahne announced an ipkg-based new version of dsmod about 2 years ago (it is still in this thread). Unfortunately, little has happened since then. The last commit to the ds-mod subversion repository on Sourceforge was 8 months ago. Meanwhile, freetz has developed in a different direction in some ways.
I think it would be sensible to consider trying to step forward to implementing dynamic packages (and opkg appears to be the best choice for this at the moment) in freetz, and based on freetz. Before doing so, however, I would suggest contacting Danisahne about his opinion and the status of his code. If I remember correctly, he also had other goals with his new version aside from ipkg. Maybe some code of his version could be used.
Anyway, I think getting this into freetz-1.1 is quite unlikely. If we decide to try implementing this, I would really consider a separate development branch (and freetz-future is just a milestone for tickets/feature requests not already assigned to a real milestone/version, it's no sandbox and there will never be a branch or release for this - so if you want to open a ticket with this, it's the right place to go. I think a ticket is a better place for discussing new features anyway).

Nico

[johnbock]

Understood.