On Tue, Apr 13, 2010 at 1:57 AM, Alexander Malysh <[email protected]> wrote: > Hi, > > Am 02.04.2010 um 16:12 schrieb hisham malik: > >> Attaching more patches for the kannel source. >> >> 1. Limit SMS Sending Retries: >> Limit SMS sending retries in split/long messages > > this is not a proper fix because you end to failing the whole multipart > message if one part can't be send > e.g. due to the throttling error and this patch doesn't handle resend via the > same SMSC.
Your observation is correct, but IMHO is not the common case (and system should be optimized for common case). In our usage, multi-part message almost always fail almost always fail when user's mailbox becomes full and cannot accept anymore messages. In this case, the existing kannel code keeps on retrying the failed multi-part messages without respecting the 'sms_resend_retry' parameter. The situation aggravates when there are multiple multi-part messages that failed due to the mailbox full issue, and the overall throughput of kannel starts to decrease since these failed parts are getting priority over new messages coming in (due to recursive nature of the code), and the userbase starts experiencing delays in message delivery. So given the existing architecture of kannel, failing the whole multi-part message and not resending via the same SMSC is a very small price to pay for optimizing for the common case, and ensuring that bad messages do not stall the servicing of rest of the messages. > >> >> 2. Handshake Error and Garbage Character in Long Message: >> Added send_failed call to the message sending function when error is >> received during handshake. >> Fixed garbage characters in multi-part messages. > > This patch is partially commited. I don't understand part about garbage > chars. Could you please explain in details > why we need this patch? If we need this patch then len calculation is > somewhere broken. > I created this patch almost a year back, and can't seem to find myself how the garbage chars was handled. Basically the issue was in the case of multi-part messages. There were some bad characters that were read out when converting from 7-bit encoding. A patch was added to remove those garbage characters. As for the other 'handshake error' part (which is part of the submitted patch), there was an else case missing which would handle an error response sent out when the initial command is sent to the handset. > Please don't use // as comment because it not valid C and if you make patches > please make it with ignore white spaces > because it's very hard to read... > My bad. I'll create the patches to be more readable next time around. Thanks. > Thanks, > Alexander Malysh > >
