vkernel.conf ?

[Chris Turner <c.turner@xxxxxxxxxxxxxxxxxxx>]

thread renamed appropriately..

Simon 'corecode' Schubert wrote:
> -1000.  getcap is a total misdesign.  you can't use colons in values and
> it is not really humanly readable.

bah!

I thinkgetcap is disliked because *termcap* settings are so terse..
Compare /etc/gettytab with /etc/login.conf..

Gettytab is very spartan and not understandable (e.g. what's a :uc: ?),
but /etc/login.conf is very clear.. (oh.. :autodelete=30d: )

I might be wrong about your preference here though..
and yes, the colon thing is a minor limitation.

But the parser is 'built in'..

Unless there is an ini parser in the system I don't know about..

> 
> what for do you need that config anyways?
> 

I think the general idea (from this and some previous discussions)
is to make vkernels completely config file based.. for example:

$ vkernel [start|stop|reset|check|console|list] [configname|all]

where configname would be an entry in /etc/vkernel.conf with
something like:

configname:\
    :memory=32m:\
    :vkd0=/path/to/file:\
    :vke0=auto,bridge0,10.0.0.1,0xDEADBEEF:\

etc.

or as you prefer:

[configname]
   memory=32m
   vkd0=/path/to/file
   ...

Note that the config file entries look just about identical...

A dhcp / named 'C' style config syntax would be my preference from a
user perspective, but definitely not from an implementers perspective..

Similarly, branching the logic for amd's 'fsinfo' parser might
be a good fit user-wise, but would be more work too..

also,
this probably brings up the 'default vkernel installation path' again..
unless 'vkernel' becomes a wrapper program around some kind of
'libexec/vk.configname' mechanism.. which actually sounds worth thinking
about..

- Chris