[Devel] db schema status, questions

Stephen Varga varga at zbzoom.net
Sat May 26 22:18:26 CEST 2007


Henning Westerholt wrote:
> On Dienstag, 22. Mai 2007, you wrote:
>   
>>> during my work with the database schemas i've noticed and fixed some
>>> issues with the current schema:
>>>
>>> - addition of a version field to pdt table
>>>       
>> Then the pdt module should also check the version number.
>>     
>
> Hello Klaus,
>
> ok, that make sense. I'll look into this next week.
>
>   
>>> - change uri: uri_user field length from 50 to 64 (standard user data
>>> length)
>>>       
>> Is this a global #define or is it hardcoded in every table definition?
>>     
>
> Of course a #define. :-) All these values are in db/schema/entities.xml 
> defined.
>
>   
>>> - change sip_trace: callid and :status from 254 to 255, and sip_trace:
>>> [..]
>>> - presentity: domain from 124 to 128 increased
>>>       
>> There is also mismatch between column types, maybe you can take a look
>> at this too.
>>    Postgresql:
>> presentity.body is a "bytea"
>> xcap_xml.xcap   is a "text"
>>     
The presentity.body column type was changed to "bytea" because the 
presence module did not work when it was set to "text". When the column 
type is "text" you will get this following error 0

(14253) PRESENCE:agregate_xmls: ERROR while parsing xml body message

The details of the issue and my solution are documented in Bug # 1684821

> This is already fixed in the new schema.
>
>   
>>> - xcap_xml: user from 66 to 64 decreased
>>> - pua:watcher_uri and pres_user increased from 128 to 255 (std. URI
>>> length)
>>>
>>> In my opinion it's more resonable to have some standard length for
>>> datatypes, what do you think about this changes?
>>>       
>> sounds reasonable
>>
>>     
>>> At the moment the IDs for the tables are either signed or unsigned.
>>> Should i change this to only one type?
>>>       
>> I think unsigned sounds better for an ID, but AFAIK postgresql does not
>> support unsigned integer. The "serial" is some kind of unsigned integer.
>>     
>
> Ok, good. Postgresql uses already "INTEGER(10) SERIAL NOT NULL" for the id 
> fields.
>
> Henning
>
>
> _______________________________________________
> Devel mailing list
> Devel at openser.org
> http://openser.org/cgi-bin/mailman/listinfo/devel
>
>
>   
Steve



More information about the Devel mailing list