Re: [Cerowrt-devel] Cerowrt-devel Digest, Vol 26, Issue 49

Re: [Cerowrt-devel] Cerowrt-devel Digest, Vol 26, Issue 49

[Mike O'Dell]

re: systemd vs procd vs etcd...

If other distros have largely converged to systemd,
is it worthwhile for CeroWrt to do something different?
This assumes that the daemons in question have already or
are in process of becoming systemd-compatible. If that is
indeed the case, is it really worthwhile to spend time
supporting something different? 

not trying to re-open old wounds, just wondering how many
different approaches are actually "better" in some material
way and how many are just "different".

I've watched Apple go through the pains of moving all the
lifetime control of services to launchd. It took a long time to
justify it being different, but now that it's done, the fact
there is only ONE place to look is really a feature. One thing,
for instance, is that the Xserver and its helpers all start
automagically when an X11 binary is run. Likewise, making
a daemon periodic instead of continuous is changed in just
one place - not moved from one to another.

My point is that making is truly better, as opposed to "just
different, yet again" requires doing the whole job, not just
a different subset of it. So if there is a base of systemd-capable
versions of the daemons in question, just use those to avoid 
the work. or do the whole job an import launchd. (which i'm
*not* lobbying for!)

      -mo

Re: [Cerowrt-devel] Cerowrt-devel Digest, Vol 26, Issue 49

[Jim Gettys]

[-- Attachment #1: Type: text/plain, Size: 2623 bytes --]

On Thu, Jan 23, 2014 at 1:07 PM, Mike O'Dell <mo@ccr.org> wrote:

> re: systemd vs procd vs etcd...
>
> If other distros have largely converged to systemd,
> is it worthwhile for CeroWrt to do something different?
> This assumes that the daemons in question have already or
> are in process of becoming systemd-compatible. If that is
> indeed the case, is it really worthwhile to spend time
> supporting something different?
>
> not trying to re-open old wounds, just wondering how many
> different approaches are actually "better" in some material
> way and how many are just "different".
>
> I've watched Apple go through the pains of moving all the
> lifetime control of services to launchd. It took a long time to
> justify it being different, but now that it's done, the fact
> there is only ONE place to look is really a feature. One thing,
> for instance, is that the Xserver and its helpers all start
> automagically when an X11 binary is run. Likewise, making
> a daemon periodic instead of continuous is changed in just
> one place - not moved from one to another.
>
> My point is that making is truly better, as opposed to "just
> different, yet again" requires doing the whole job, not just
> a different subset of it. So if there is a base of systemd-capable
> versions of the daemons in question, just use those to avoid
> the work. or do the whole job an import launchd. (which i'm
> *not* lobbying for!)
>
>       -mo
>

Mike,

CeroWrt is an upstream development version of OpenWrt.

One of the current constraints of OpenWrt is that it still (for its own
good reasons) targeted at very small flash routers (8mb, and even 4mb flash
routers).

When I last looked at systemd, it's footprint looked larger than would
likely be feasible given those constraints: granted, I did not do a really
careful analysis of systemd's footprint, which is probably only knowable by
doing a full port.

So for the moment, I expect our attention needs to be elsewhere (though I
do like systemd from what I've seen of it).  But without funding to bear
the costs of such a fork from OpenWrt, it can't fly.

I expect the flash constraints will eventually ease; the question is
when...  Funding for these projects (OpenWrt/CeroWrt) would also help, as
it would make sense to do a fork as it's clear that the flash constraint is
something that should bite the dust someday and sooner would be better than
later...

                                 - Jim



> _______________________________________________
> Cerowrt-devel mailing list
> Cerowrt-devel@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cerowrt-devel
>

[-- Attachment #2: Type: text/html, Size: 4141 bytes --]

Re: [Cerowrt-devel] Cerowrt-devel Digest, Vol 26, Issue 49

[Mike O'Dell]

great. i didn't understand the flow between 
Cerowrt and Openwrt.

carry on. (grin)

      -mo

Re: [Cerowrt-devel] Cerowrt-devel Digest, Vol 26, Issue 49

[David Lang]

[-- Attachment #1: Type: TEXT/Plain, Size: 3022 bytes --]

On Thu, 23 Jan 2014, Jim Gettys wrote:

> On Thu, Jan 23, 2014 at 1:07 PM, Mike O'Dell <mo@ccr.org> wrote:
>
>> re: systemd vs procd vs etcd...
>>
>> If other distros have largely converged to systemd,
>> is it worthwhile for CeroWrt to do something different?
>> This assumes that the daemons in question have already or
>> are in process of becoming systemd-compatible. If that is
>> indeed the case, is it really worthwhile to spend time
>> supporting something different?
>>
>> not trying to re-open old wounds, just wondering how many
>> different approaches are actually "better" in some material
>> way and how many are just "different".
>>
>> I've watched Apple go through the pains of moving all the
>> lifetime control of services to launchd. It took a long time to
>> justify it being different, but now that it's done, the fact
>> there is only ONE place to look is really a feature. One thing,
>> for instance, is that the Xserver and its helpers all start
>> automagically when an X11 binary is run. Likewise, making
>> a daemon periodic instead of continuous is changed in just
>> one place - not moved from one to another.
>>
>> My point is that making is truly better, as opposed to "just
>> different, yet again" requires doing the whole job, not just
>> a different subset of it. So if there is a base of systemd-capable
>> versions of the daemons in question, just use those to avoid
>> the work. or do the whole job an import launchd. (which i'm
>> *not* lobbying for!)
>>
>>       -mo
>>
>
> Mike,
>
> CeroWrt is an upstream development version of OpenWrt.
>
> One of the current constraints of OpenWrt is that it still (for its own
> good reasons) targeted at very small flash routers (8mb, and even 4mb flash
> routers).
>
> When I last looked at systemd, it's footprint looked larger than would
> likely be feasible given those constraints: granted, I did not do a really
> careful analysis of systemd's footprint, which is probably only knowable by
> doing a full port.
>
> So for the moment, I expect our attention needs to be elsewhere (though I
> do like systemd from what I've seen of it).  But without funding to bear
> the costs of such a fork from OpenWrt, it can't fly.
>
> I expect the flash constraints will eventually ease; the question is
> when...  Funding for these projects (OpenWrt/CeroWrt) would also help, as
> it would make sense to do a fork as it's clear that the flash constraint is
> something that should bite the dust someday and sooner would be better than
> later...

not to mention other issues like the systemd journal now writing everything from 
stdout and stderr of programs to it's logfile, with the associated flash wear 
this would cause.

Desktop distros are largely (but not entirely) moving to systemd, what embedded 
distro are you aware of that has moved?

Also, what Enterprise distro has moved? RHEL 7 is going to ship with systemd, 
but since it hasn't shipped yet, we don't know what it's acceptance is going to 
be in the enterprise space.

David Lang

[-- Attachment #2: Type: TEXT/PLAIN, Size: 164 bytes --]

_______________________________________________
Cerowrt-devel mailing list
Cerowrt-devel@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/cerowrt-devel