On 24.08.2009 15:57 Uhr, Iñaki Baz Castillo wrote:
2009/8/24 Daniel-Constantin Mierla miconda@gmail.com:
I am personally aware of companies using Kamailio with several millions of subscribers, using kamailio database schema. Also, I am aware of companies having more or less same level of subscriber base using SER database schema. All have additional tools for management, integration with third-party application, a.s.o. Do you think that saying "hey, you were the unlucky bastard because we are going to drop tomorrow the database schema you are using" is the solution?
That's right, but I don't expect that migration to SR is a priority for those companies already using OpenSER/Kamailio. For example, I know some companies still using OpenSer 1.2.
yes, I do, even older, I have a ser 0.9.4 somewhere and no plan to upgrade it. It has 1 year 4 months without restart.
However, it seems that the idea is that Kamailio/SER will use SR as core, but the fact is that SR core uses current Kamailio and SER modules/features as separate (modules_k, modules_s...).
some modules were moved under "modules" directory. The ones that have names overlapping and inter-dependencies are going to stay in separate directories.
Wouldn't make sense to unify SR code instead of having it splited in K and S?
probably you misunderstood something. The code is unified, but SR has support to hold modules in more than one directory. Now the structure is based on origin (again, the reason is name overlapping), but in the future could be:
modules_presence modules_db
IIRC, there are over 150 modules all together, so better structuring might be necessary.
AFAIK this is the idea for a future, but in the meanwhile I don't fully understand why SR requires to run right now as it is.
What I don't understand is the reasons to make current SR working with K and S features/modules compatibility. We don't need a SR working solution right now (since Kamailio and SER do exist), do we?
Maybe not you, but there are others. I am facing many troubles because of TCP (also TLS) layer in K which do not happen with SR core - asynchronous TCP helps a lot.
Sure that's a good reason :) But again you are speaking about Kamailio using SR as core (new and improved TM module) while I meant SR code itself (which for now is a mostly separate mix between K and S code instead of an unified technology).
I don't get it. Maybe you can rephrase/detail what is misleading/confusing you. The existence of 3 module directories? Everything is unified, one source repository, one project, if you think just to SR environment.
All (but seas) are ported.
Perhaps I understand it wrongly, but IMHO the modules will be really ported when there is a unique MI interface for all of them (instead of having to use K MI for K modules and S MI for S modules),
Using RPC interface gives access to MI commands -- so ser guys are the first lucky :-). However, let's get back to the root: - there are modules implementing MI - there are modules implementing RPC interface - there are modules implementing none - there might be modules implementing both
Where to position a module is just a matter of developers. Similar choise is for regular expressions - use posix or pcre library - database - use interface v1 or v2 (which has prepared statements support) - and examples can continue. I see ability to choose a great feature.
when there are just an unique class of pseudovariables (instead of having K and S pvs)...
Here is only one: config file variables - how they are referred in mails is a different story, to better indentify and reflex origin, but all can be used in config file. Actually, the group referred K PVs have classes.
Cheers, Daniel