Stephen Hahn wrote:
> * Rinaldo DiGiorgio <[EMAIL PROTECTED]> [2008-04-27 22:15]:
>   
>> Dan Price wrote:
>>     
>>> On Sun 27 Apr 2008 at 12:54PM, Rinaldo DiGiorgio wrote:
>>>   
>>>       
>>>>>> [EMAIL PROTECTED]:~# cat /etc/motd
>>>>>> Sun Microsystems Inc.    SunOS 5.11    snv_86    January 2008
>>>>>> [EMAIL PROTECTED]:~# grep 11 /etc/motd
>>>>>> Sun Microsystems Inc.    SunOS 5.11    snv_86    January 2008
>>>>>> [EMAIL PROTECTED]:~# echo $?
>>>>>> 0
>>>>>> [EMAIL PROTECTED]:~# grep 12 /etc/motd
>>>>>> [EMAIL PROTECTED]:~# echo $?
>>>>>> 1
>>>>>>     
>>>>>>         
>>>>>>             
>>>>>   The analogy here isn't convincing, possibly because the pkg(1)
>>>>>   invocations above are a small subset of those available.  pkg
>>>>>   uninstall of an unknown package gives a non-zero exit status; pkg
>>>>>   install of an unknown package gives a non-zero exit status.  At the
>>>>>   completion of the pkg install invocation in both the cases above, the
>>>>>   requested package is actually installed.
>>>>>           
>>> A similar command that you might use to make this test is
>>> 'pkg list <pkgname>' -- this does indeed return 1 if the package
>>> is not installed.  I think that's the most analogous to grep, and
>>> is the test we use in the test suite for this kind of checking.
>>>   
>>>       
>> Excellent, thank you. I have put together a JSON interface  available
>> to any HTTP capable client  for remote installation of packages. I
>> typically manage about 100 systems.  The ability for me to install
>> from one server will save me quite a bit of time.  I had something
>> similar with the old packaging system but the commands were too
>> verbose and asked too many questions. So far the new system is much
>> better for me on this issue.
>>     
>
>   So, since you're one of the first people to attempt to automate use of
>   pkg(1), what output would you like to see, or to see omitted?  (We are
>   in complete accord about "no questions".)  Dan has already made the
>   output sensitive to whether it is outputting to a tty or not, but
>   further output modifications are certainly possible.
>
>   - Stephen
>
>   
 I noticed that the pkg software was aware of a non tty access path and 
did not
send screen control commands.  Your output is currently easy to parse 
using regex.
Many people will consider a regex approach a hack(negative sense) so it 
would be great
if you considered  a standard machine  syntax or had some guarantee that 
output changes
 would be managed. I have  seen some hints that  you  already have   
machine readable
output but I did not  look into that yet.  The two  formats that I see 
most often are JSON
and XML when communicating with HTTP wrappers providing a RESTful 
interface to a
remote service. It is not clear that it is  the right approach for CLI 
commands. I personally
would like to see every Solaris command  have an option for JSON or XML 
output.

Rinaldo







_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to