Brain Dump #1

Joe Armstrong <>
Thu Feb 6 15:04:18 CET 2003


On Thu, 6 Feb 2003, Mickael Remond wrote:

> * Vlad Dumitrescu (EAW) <> [2003-02-06 13:06:26 +0100]:
> 
> > - Instant messaging & NoXML & pure Erlang communication: maybe it won't be a bad idea to have an interface towards the other IM systems, too - let's leave the door open for the world. Also, I don't think you mean to throw away the existing protocols (that work), just to make them use another encoding?
> 
> The fusion of the Jabber, Zope based on pure Erlang communication is a very
> good idea. Thierry and I are thinking that wa can indeed do better that those tools.
> Zope is a good product regarding end user functionnalities and scope but is
> relying on a poor design technology compared to Erlang.
> 
> However, I did not grasp your transition between this first idea and the fact
> that we need volunteer to parse true type font and could handle PDF
> I do not see this as the most direct way to implement this kind of platform.
> So why do you see this part as the most important things to begin with ?

  It's not the most important to  begin with - this was a *brain dump*
- this is the problem that interest me most at the moment.

  From the point of view of a  community project a set of servers in a
loose peer-to-peer network with a plug-in architecture is what I think
is needed to boot things off.

  I wrote a skeleton server with this  in mind a while back - I'll dig
it out and write it up as  a tutorial - If we could add the leadership
election of Uffe + Thomas Arts we'd be in business ....

  question: how many people have machines with permanent un-firewalled
Internet access? - I  have a few at SICS - we  need these to bootstrap
the system....

  I think any community project needs  a lot of *writing* doing - most
community projects seem to me to be useless because I can't understand
what they do and  what they are - the people *in*  the project all get
wildly  excited -  they  hack code  like  mad and  nobody outside  the
project can understand *anything*

  We must not fall into this trap.

  We need clear architectural documents - written in plain no-nonsense
English that everybody can understand  - otherwise we'll just get "yet
another bright idea that failed"

  A lot of the computer stuff we see around us is just plain junk - if
we want to  improve this we have to explain  pedagogically *why* it is
junk and then show how it can be improved.

  If people do not understand something they will not use it.

  /Joe
   

> 
> > - send mail to things: isn't this basically the same idea as behind Jini or
> > WebServices, for example? Just a matter of protocol and of scale :-)  I think
> > it can be done and should be done, and Erlang is the best tool!
> 
> Regarding Jabber, I think it is not very different than an SMTP based messaging
> system. Jabber is nothing more than a logical bus based on XML. I have thought
> on several design for such a bus and at the end I think that it always come
> very close to an SMTP server. The most simple way to build such a data oriented
> bus is to build it arount an SMTP server.
> 
> However, some specific element of the Instant Messaging protocol such as the
> "presence" indicater needs something different (Connexion to the server is
> kept).
> 
> 




More information about the erlang-questions mailing list