boot ordering and syslogd

[Sam Leffler]

Doug Barton wrote:
> Sam Leffler wrote:
>> But saying "it's not a problem that needs to be fixed" is utterly lame.
>
> Since that's not what I intended to convey let me be less concise.
>
> 1. Changes to the current rc order need to have a very solid and well 
> thought out rationale which includes a detailed cost/benefit analysis 
> since historically it's been difficult to anticipate all of the 
> potential fallout from a change, and we don't want to violate POLA.
>
> 2. You are stating a problem case whose cost does not sound compelling 
> to me, both because I personally don't think "fixing" it would provide 
> much value, and because it would cause a significant change in the 
> current order which inevitably brings a great deal of cost (whether 
> those costs can eventually be ameliorated or not).
>
> 3. The burden is on the one requesting a change to demonstrate its 
> benefit. Saying "it's a problem and needs to be fixed" (not 
> accompanied by patches) could also be considered "lame." :)
>
> 4. My analysis of the potential costs and benefits might well be 
> wrong, which is why I'd like some other people to weigh in.
>
> To summarize, I am not saying I don't think it's a problem. I'm saying 
> that we need more discussion to determine what the problem actually 
> is, what solutions are available, and at what cost.

I have _already_ demonstrated a need--the way things currently are ANY 
privsep/chroot application started prior to syslogd will NEVER log any 
messages.  I noticed this because of dhclient (and wpa_supplicant which 
very shortly will be logging msgs to syslog) but I expect there are 
other services if you look and I wouldn't be the least surprised if more 
get added (does named work?).

I didn't expect a fix to be simple; if it were I'd have just done it 
myself.  Saying "fix it your self and send us a patch" is worse than lame.

    Sam