I'm writing a server that is specifically designed to manage weblog
posts for multiple weblogs.
I have a C++ class called Blog. It pretty much encapsulates a single weblog.
Currently, I have a separate c4_Storage that is initialized as a
member of the Blog class. That is, each blog will have it's own
metakit file on disk. I anticipate between 1-2k posts before
archiving becomes a concern for each individual weblog. The metakit
structure is very simple, "posts[id:S,updated:S,msg:S]" where the
'msg' column holds the actual content as XML. I also have a separate
validation class that uses Metakit to store auth information in yet a
separate metakit file.
Are there any efficiency concerns with this? Would I be better off
having a single metakit file with "posts" as a subview? Are there
concerns over the overhead for each file?
Since the blogging system is going to be distributed (in the sense of
peer2peer), I anticipate high traffic requesting content as well as
exchanging post data between different nodes. The first time a
request is made, either to enter a new post, or to retrieve the
contents of a post, the Blog class is instantiated for that weblog.
Commits are made on every validated change.
I anticipate that the bulk of the time will be waiting for I/O
operations over the network, but I want to make sure there isn't some
"gotcha" with having 5 or a dozen metakit files open and commiting to
them simultaneously.
Anyway,
Thanks for the time to read this rambling email....
Darel
_____________________________________________
Metakit mailing list -
Metakit@...
http://www.equi4.com/mailman/listinfo/metakit
