Re: [Agendaware-dev-server] Re: Doc

[Shane T. <tur...@ya...>]

Yea, I'd like to have a look at the code too before we
submit it to the source tree to see if we're all on
the same page.  I'd also like it (if everyone agrees)
to have some sort of designs put together first to
show exactly what we intend to do before jumping into
coding too much plus give each other a chance to make
suggestions or maybe catch holes in the designs.  I 
submitted an archictecture layout if everyone would
take a look at it and see if it makes sense of if you
have suggestions or changes you think we should make.
Proper design is going to be crucial upfront to make
this project work.

Shane
--- zack <za...@th...> wrote:
> Hey Agyani.
> 
> Thanks for the email.  Comments inline.
> 
> zl
> 
> 
> Agyani wrote:
> 
> > Hi,
> > 
> >  
> > 
> > Tests are under way with the server. It has
> started to accept connection 
> > on port 7007 and is presently printing every input
> from the client on 
> > the stdout. I am working to make the ssl stuff
> running.
> > 
> >  
> 
> 
> 
> Shoot me the code for that when you can.  If all you
> have so far is 
> non-ssl, I'll start with that.
> 
> 
> > 
> > Gone through the doc, but I have not gone through
> the rfc mentioned 
> > (will try to go through). Your comments were
> appropriate for the 
> > iCalendar interface as nothing can be said without
> understanding the 
> > whole protocol.
> > 
> >  
> > 
> > I agree with Shane, we do require loadbalancing
> functionality, but we 
> > can have a load balancer built as a standalone
> application that can 
> > provide loadbalancing for the servers and not to
> integrate this feature 
> > in the server itself. This proposed loadbalancer
> can  have a config file 
> > which stores the information for available servers
> and then 
> > distributes incoming connection to them. If we are
> able to make the 
> > loadbalancer totally abstract form the underling
> server then this can be 
> > used with any other server
> 
> 
> An easy way to do this would be when logging into a
> server (you could 
> have primmary and secondary configged on the client)
> the server then 
> assigns you a server to use throughout the session. 
> This option still 
> leaves the database server as a single point of
> failure though, and you 
> couldn't really run a db on each server as you would
> still have to have 
> a single database.  So, the database would have to
> be clustered (beowolf 
>   maybe?).  If you have to set up a clustered
> database, it might be 
> worthwhile looking into what it would take to make
> the agenda server 
> cluster aware.  So, I guess we need to look at what
> is possible, and 
> what level of redundancy/load balancing it really
> gives us.
> 
> 
> > 
> >  
> > 
> > For connection distribution between servers for
> different functionality 
> > we can bind the loadbalancer's socket to all the
> addresses of the 
> > host(this may not be a good idea) and for each
> incoming connection the 
> > loadbalancer checks the address and distributes
> accordingly.
> > 
> > The proposed server is going to have a config file
> having various config 
> > information i.e port,ip, log etc. so to implement
> the above won't be a 
> > problem.
> > 
> >  
> > 
> > XML parsing is mostly done on document basis i.e
> the whole document is 
> > constructed and send for parsing as we require all
> the information in 
> > the document before we can act on it. So it would
> be better if we lean 
> > toward the document way.
> > 
> 
> 
> Sounds like DOM is the way to go.
> 
> 
> >  
> > 
> > I agree the client should have the functionality
> to do all his work 
> > offline and when he wants to update the server he
> simply connects to the 
> > server and updates/sends the information.
> > 
> >  
> > 
> > Some message archiving features could also be
> good.
> > 
> 
> 
> Good.  I think any archiving should be done on the
> server.
> 
> 
> >  
> > 
> > The proposed server will be more towards c++ then
> c, this gives us the 
> > upper hand in implementing class modularity and
> abstraction. Taking the 
> > case of db, the class that will provide database
> access will have a std 
> > set of methods for getting,setting,updating and
> retrieving information 
> > from the db. These std set of methods will be
> defined in a abstract 
> > database class having all pure virtual methods.
> Now the question of 
> > using other databases can be answered easily, to
> make the server use a 
> > new database the implementar has to extend the
> base class and provide 
> > the rest of the functionality, recompile the
> server and you have it 
> > working. I propose some thing similar to JDBC
> classes and their methods.
> > 
> >  
> > 
> > I don't have a static IP so my box can't be used
> as a server. You can 
> > have linux setup on you system and use it as the
> server and client both. 
> > I will be uploading all the information for
> compiling,configuring and 
> > running the server and the sql to setup the db
> with the release. Also 
> > thinking of the db structure and will come up with
> some tables and views 
> > asap.
> > 
> 
> 
> Very cool.
> 
> 
> >  
> > 
> > Well that's it, just got your new mail Re: LDAP
> will reply shortly.
> > 
> > You mentioned you needed help on the client, what
> kind of ?
> > 
> 
> 
> Yeah, right now, I am working on it, and one other
> is joining, Sandeep. 
>   We'll be doing the client, unless someone else
> wants to help.
> 
> 
> > Has every body joined the mailing list ?
> > 
> 
> 
> Everyone who has commited to the project is on the
> mailing lists.  There 
> are still a couple of people who have emailed me,
> but I am not sure they 
> are commited.
> 
> 
> >  
> 
=== message truncated ===


__________________________________________________
Do You Yahoo!?
Find a job, post your resume.
http://careers.yahoo.com