tomcat-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Cris Perdue <c...@perdues.com>
Subject Re: request for review: server/config discussion
Date Thu, 22 Jul 1999 20:47:13 GMT
A lot of the discussion has centered on how to design the server to
respond to changes in a config file or similar thing. Another equally
important point of view, if not more so, is the _internal_, _running_
configuration of the server, the actual internal state.

In other words, no matter how it received a description of the desired
setup, at all times the server internally has an actual, current
configuration.  It has a certain "Document Root".  It has a certain set
of ports it listens to, a certain number of threads doing certain
things, session timeouts, etc. etc..

No matter how good the mechanism for receiving updates to a config file,
dynamic configuration of the server can only be as good and as flexible
as the APIs it has internally for updating its running state, and the
implementation of these APIs. These mechanisms (of course) are specific
to Web servers in general and even specific to the design of the
particular Web server.

To take this subject and run with it a bit further, it's been my
experience that properties (like JavaBeans properties) and collections
with insert, delete, and setting the entire list, plus perhaps
associative arrays, are abstractions that work well with config files
and can handle a lot of application needs.

Personally I believe if you can build a server with clean, dynamic,
internal reconfiguration, building mechanisms for responding to changes
in config files shows up as a relatively straightforward engineering
task. And if the internal interfaces are not available, you can't
dynamically reconfigure at all.
--
Cris Perdue
Impact Online, Inc.
http://www.volunteermatch.org



Mime
View raw message