Alex,

Recently i had the case that a DLR were wrongly updated.

2005-12-21 08:00:11 [2436] [16] DEBUG: DLR[mysql]: Looking for DLR smsc=SMSC-1, 
ts=211205080007, dst=<masked>, type=2
2005-12-21 08:00:11 [2436] [16] DEBUG: sql: SELECT mask, srvc, url, src, dest, 
boxc FROM dlr WHERE smsc='SMSC-1' AND ts='211205080007';
2005-12-21 08:00:11 [2436] [16] DEBUG: sql: DELETE FROM dlr WHERE smsc='SMSC-1' 
AND ts='211205080007' LIMIT 1;

2005-12-21 08:00:15 [2436] [16] DEBUG: DLR[mysql]: Looking for DLR smsc=SMSC-1, 
ts=211205080007, dst=<masked>, type=1
2005-12-21 08:00:15 [2436] [16] DEBUG: sql: SELECT mask, srvc, url, src, dest, 
boxc FROM dlr WHERE smsc='SMSC-1' AND ts='211205080007';
2005-12-21 08:00:15 [2436] [16] DEBUG: sql: DELETE FROM dlr WHERE smsc='SMSC-1' 
AND ts='211205080007' LIMIT 1;

On a hudge systems, ts are in many case the same as many MTs (on a same SMSC).
Do you really think that most of DLR should be wrong just because of a few SMSC 
thats are returning wrong ADC ?
IMHO, it does not break all existing installations as you may think, it will 
just not found the DLR into the DB... 
but i think this is really acceptable.

Vincent.


----- Original Message ----- 
From: "Alexander Malysh" <[EMAIL PROTECTED]>
To: <[email protected]>
Sent: Tuesday, January 10, 2006 12:10 PM
Subject: Re: [PATCH] DLR with field_dst not supported in Mysql/Pgsql


> Dear Vincent,
> 
> did you find the previous discussion about this issue? If not please 
> find it in ML archive read carefully and then make your statements.
> 
> The problem is simple. UCP/SMPP does not define how destination address 
> should be returned in delivery report. It does not explicitely say: it 
> _must_ be equal. So you have no right to say those SMPP/UCP servers are 
> buggy (ok, I would say it too, but...). And we (developer) have no right 
> to break existing installation w/o a good reason and yes I see good 
> reason to use destination in select, update, delete but not so how you 
> do it. Destination should be stripped for some chars at beginning and 
> only then used in select/update/delete.
> 
> Thanks,
> Alex
> 
> P.S. Please also see my patch for this topic in archives.
> 
> 
> Vincent CHAVANIS wrote:
>> Dear,
>> 
>> Alex i understand your point of view,but it's better to not found a DLR in 
>> the DB than updating a wrong one... (safety)
>> It breaks compatibility with broken UCP servers that allow to change the 
>> OAdC field, YES
>> But, many UCP servers does not allow this. SMSC should return the AdC as 
>> mentionned in UCP 51
>> That's why i think that this feature should be implemented.
>> 
>> regards
>> 
>> Vincent
>> 
>> --
>> Telemaque - NICE
>> Service Technique
>> Tel : +33 4 93 97 71 64 (fax 68)
>> [EMAIL PROTECTED]
>> 
>> ----- Original Message ----- 
>> From: "Alexander Malysh" <[EMAIL PROTECTED]>
>> To: <[email protected]>
>> Sent: Tuesday, January 10, 2006 11:33 AM
>> Subject: Re: [PATCH] DLR with field_dst not supported in Mysql/Pgsql
>> 
>> 
>> 
>>>Hi Stipe,
>>>
>>>why did you commit this pacth w/o waiting a bit for comments?
>>>
>>>such patch was already discussed few times and we did not found solution 
>>>yet. I'm -1 for this patch because it will break too many installations.
>>>For short explanation why -1: if you sent destination 00491234567 but 
>>>SMSC returns 01234567 as destination (IIRC this is the case T-Mobile 
>>>germany UCP server) then you will not find DLR in the DB. For a longer 
>>>explannation seek in ML archives for discussion about it.
>>>
>>>Please revert it or I will do it.
>>>
>>>Thanks,
>>>Alex
>>>
>>>Stipe Tolj wrote:
>>>
>>>>Vincent CHAVANIS wrote:
>>>>
>>>>
>>>>>Dear,
>>>>>
>>>>>I've noticed today that field_dst is not implemented in DLR for mysql 
>>>>>and pgsql.
>>>>>If a DLR arrives with the same TS as another one on a same SMSC, 
>>>>>kannel will not be able to update the correct one
>>>>>
>>>>>Here is a patch tested and in production status, maybe should be 
>>>>>commited to CVS
>>>>
>>>>
>>>>Hi Vincent,
>>>>
>>>>yep, ++1 here. Commited to CVS.
>>>>
>>>>A kind request, please do not cross-post to both lists, and we'd prefer 
>>>>'diff -u' format patches ;)
>>>>
>>>>Stipe
>>>>
>>>>
>>>
>>>
>>>
>> 
>> 
>> 
> 
> 
>


Reply via email to