|
My two (oops!
three) cents:
1. A GUID or UUID (Globally unique
ID) can be generated. Libraries are available to generate such a unique ID
depending on time or a random number on Linux.
Pros: Always guarantees unique ID
irrespective of number of bearerboxes or message submission rate.
Con: Platform specific
code.
2. Another ID generation can
be similar to SMSC standards for generation of message IDs which is
<timestamp>:<destination number>.
Pros: Easy and platform
independent.
Con: This is a subset of Stipe's Heuristical msg id generation.
And also here most of the SMSC take timestamp to be of seconds granularity. Thus
binding on gateways to send only one message per second towards a particular
destination. So we can think of keeping a finer granularity approaching Millis.
Also, under rare circumstance when multiple bearerboxes are running, it doesn't
guarantee unique message ID across boxes.
3. One of the solution can be to assign unique box id to each bearerbox
running in the system and have message ID as <timestamp in
Millis>:<destination>:<box id>
comments?
Jeetendra.
----- Original Message -----
Sent: Friday, August 22, 2003 5:13
PM
Subject: [RFC] unique SMS IDs
Hi list,
currently bearerbox acts like:
whenever an
SMS is passed to the store-file it get's an incrementaly
counted ID and
then stored.
It is desirable (for several purposes) to have unique IDs
for messages
coming in (SM-MO), hence when entering the system from SMSC
side and
when leaving to the HTTP (application server) side, and of course
vise
versa.
Hence, I'd like to discuss how we can acchive uniqueness
of IDs
accross even several parallel running bearerbox instances on one
box.
>From our perspective there are two ideas:
*
heuristical: assume that a constructed ID = timestamp + sender +
receiver +
counter is unique accross all instances.
* ID pool caching:
which means there is a thread in bearerbox with a
ID queue. The queues in
several concurrent bearerbox processes are
distinct, which means the thread
has to garantee whereever he obtains
a pool of IDs, that those are assign
only to himself.
Option 1 is of course simpler to do. Option 2 is
*stronger* in the
sense of high-load systems and consistancy.
Any
thoughs for this?
BTW, I'd also like to add the SMS ID to the
access.log. Any
objecttions here?
Stipe
[EMAIL PROTECTED]
-------------------------------------------------------------------
Wapme
Systems AG
Vogelsanger Weg 80
40470 D�sseldorf
Tel:
+49-211-74845-0
Fax: +49-211-74845-299
E-Mail: [EMAIL PROTECTED]
Internet: http://www.wapme-systems.de
-------------------------------------------------------------------
wapme.net
- wherever you are
| 