Re: Bug#19133: distributed-net: support PPP -- /etc/ppp/ip-up.d/distributed-net wanted
[You (Manoj Srivastava)]
> Firstly, I think we should minimize the number of conffiles on
> the system. A one line addition to the script shall meke this
> unnecessary.
Ok, well, I see your point. I still don't agree and I think a uniform
administrative GUI would be a better solution. I see uniformity of
operation as a goal for the distribution also, and the current
/etc/ppp/ip-{up,down}.d scheme has uniformity with "unwritten policy",
that is to say, with what I call our general "run-parts" system. So
we're leaning on the system administrator's familiarity with "scripts in
a directory".
However I encourage you to try to raise a consensus and get your scheme
approved. I'm in "status quo" mode until after the freeze. ;)
> Secondly, I do not like the concept of changing the operation
> of a package, having it send off packets off machine without
> authorizzation, and tell the users change permission on the script if
> you don't like this.
>
> Changing permissions is a hack, which is required only because
> of lack of foresight in the script.
?? You call it a hack, I call it the "mode of operations" for our
existing, consistent, "run-parts" scheme. It harks back to the old,
venerable SysV init scheme.
>Adam> If you want it shipped turned off, then ship it
>Adam> without the execute bit.
>
> This is a hack. No one should ask users to go about changing
> permissions on programs, especially when a simple solution exists.
What do you do, Manoj, when you want to stop xdm from running at boot? I
don't know about you, but I start 'chmod'ing.
>Adam> Why add more debian-specific, PPP-specific infrastructure for
>Adam> well-established 'run-parts' type infrastructure.
>
> Oh, for gods sake. No one is changing "run parts
> infrastructure". We are asking for a standard control infrastructure
> for ip-up/down scripts, as running all scripts at all times is sub
> optimal.
I think this is overkill, because 90+% of users are desktop home
machines, and
> Additionally, we should minimize conffiles, the solution
> offered here provides a single point of control, the parsing
> is easy, the syntax of the conffile is simple enough that GUI
> frontends can be provided, if needed, and no one has to hack scripts
> or permissions to turn an ip script on or off.
>
> I class the current situation as conffile abuse ;-)
Um, but you're adding one additional conffile and not removing any!
>>> The default should be OFF, since nothing is put in the file at
>>> all. If the sysadmin wants it, they can edit in the configuration
>>> file.
>
>Adam> See above. Certainly possible I am not adversed to shipping it
>Adam> mode 0644 rather than 0755....
>
> Hack. Hack. Hack.
Your opinion. Works for /etc/init.d, for /etc/cron*.d, for
/etc/cfgtool.d (?), stretching imagination for a bit, for the new
/etc/skel proposal, for the new emacsen-common scheme. I think it's
elegant, uses the file system as a database rather than some file with
arbitrary grammar.
I think the filesystem makes a damn good database in linux, BTW, and I
don't mind using it that way.
> Any site worth its salt already handles all the stuff that
> ip-up down scripts do. That is what I was referring to.
Why shouldn't our distribution ship already doing what is reasonable for
the greatest number of users? If we polled to see how many people using
PPP links are currently flushing outbound mailq on link up, I would
expect that you would find it is less than 25%.
> Again, this is a general proposal for all ip-up down scripts,
> and one which provides an easy oevrview of all script active
> (cat /etc/ppp/ip.conf)
I guess we got a plain, old difference of opinion. I think
'ls -F /etc/ppp/ip-up.d/' is easier and going to be more intuitive.
.....A. P. Harris...apharris@onShore.com...<URL:http://www.onShore.com/>
Reply to: