Allow configurable 'nobody' user

[Mladen Turk <mt...@apache.org>]

Hi,

The patch allows to set the hard coded 'nobody' user
to something that eventually exists on the box
(added --with-user and --with-group (defaults to --with-user)
to configure.

If not specified they default to 'nobody' as before
Not sure about naming since I'm not very clear for what that
user is used at the first place, so I didn't open the Jira directly.


Regards
-- 
^TM

Re: 64-bit sizes in VIO and API for 2.0 ?

[Paul Querna <pa...@querna.org>]

On Thu, Dec 17, 2009 at 9:09 AM, John Plevyak <jp...@acm.org> wrote:
>
> Currently, all the sizes in the VIO/VConnection and corresponding InkAPI
> are all
> 32-bit (int on x86_64).
>
> Do we want to change these to 64-bit before 2.0 so as to minimize API
> changes
> for 2.X?
>
> Pros:
>  - reduce interface changes going forward
>
> Cons:
>  - actual large document support many require additional changes
> which we may not want to do for 2.0.
>  - in practice large documents are unreasonably inefficient in 2.0

As an open source project, I'd vote just do it, I believe there is a
general agreement we should support large files, so the interface
changes would seem inevitable, just a matter of time/hacking to
actually support it underneath.

I wouldn't worry about efficiency -- just fixing all of the APIs as
much as possible is a great start, we can optimize and change the
internals as needed in 2.0 or .2.x.

64-bit sizes in VIO and API for 2.0 ?

[John Plevyak <jp...@acm.org>]

Currently, all the sizes in the VIO/VConnection and corresponding InkAPI
are all
32-bit (int on x86_64).

Do we want to change these to 64-bit before 2.0 so as to minimize API
changes
for 2.X?

Pros:
  - reduce interface changes going forward

Cons:
  - actual large document support many require additional changes
which we may not want to do for 2.0.
  - in practice large documents are unreasonably inefficient in 2.0

Re: Allow configurable 'nobody' user

[Leif Hedstrom <zw...@apache.org>]

On 12/17/2009 11:37 AM, Mladen Turk wrote:
> I'll modify the patch so that those consts gets setup via compile
> time define (eg. -DSYS_USER="someuser").


Great!

-- Leif

Re: Allow configurable 'nobody' user

[Mladen Turk <mt...@apache.org>]

On 12/17/2009 04:06 PM, Leif Hedstrom wrote:
>
> We do need to fix the following as well, preferably with this "bug", but
> I'm ok fixing it later:
>
> ./proxy/config/records.config.in:CONFIG proxy.config.admin.user_id
> STRING nobody
> ./proxy/mgmt2/cop/TrafficCop.cc:static char admin_user[80] = "nobody";
>
>
> and not sure if we need / want this (also in TrafficCop.cc):
>
> // bitch if the admin user is not "nobody"
> if (strncmp("nobody", admin_user, sizeof(admin_user)) != 0) {
> cop_log(COP_WARNING, "proxy.config.admin.user_id should be \"nobody\"");
> }
>

Think all of this can be the same user defined via --with-user
I'll modify the patch so that those consts gets setup via compile
time define (eg. -DSYS_USER="someuser").


Cheers
-- 
^TM

Re: Allow configurable 'nobody' user

[Leif Hedstrom <zw...@apache.org>]

On 12/17/2009 01:15 AM, Mladen Turk wrote:
> Hi,
>
> The patch allows to set the hard coded 'nobody' user
> to something that eventually exists on the box
> (added --with-user and --with-group (defaults to --with-user)
> to configure.
>
> If not specified they default to 'nobody' as before
> Not sure about naming since I'm not very clear for what that
> user is used at the first place, so I didn't open the Jira directly.

Cool. These changes looks good, and I'd be ok to have this committed 
(attach to the Jira though).

We do need to fix the following as well, preferably with this "bug", but 
I'm ok fixing it later:

     ./proxy/config/records.config.in:CONFIG proxy.config.admin.user_id 
STRING nobody
     ./proxy/mgmt2/cop/TrafficCop.cc:static char admin_user[80] = "nobody";


and not sure if we need / want this (also in TrafficCop.cc):

   // bitch if the admin user is not "nobody"
   if (strncmp("nobody", admin_user, sizeof(admin_user)) != 0) {
     cop_log(COP_WARNING, "proxy.config.admin.user_id should be 
\"nobody\"");
   }



Cheers!

-- leif