First of all accept my apologies for posting in here but I can't find any other place where to post this argument.
I've playing with Scalix for months now and tried to set up everything in order to have better possible perfomance from the given server hardware. For instance you have to know I've choosen SUSE 10.1 for both stagin environments and production ones.
Many efforts has been implemented to optimize disk(s) access, to increase ram memory, to adopt faster processors, to optimize apache and so on ... anyway, not so surprisingly, performances experienced on Windows clients (especially ones with Outlook and related connector) did not match our expectations: Outlook natively connected to Exchange is far faster than the the same one connected to Scalix through it's connector. And the speed loss, in my very humble opinion, was far beyond the presence of an intermediate layer of communication (the connector itself).
Even talking about IMAP we've been experiencing poor performance on heavily populated Scalix imap folders (thousands of messages) simply during the retrieval of their headers or while downloading huge attachments. On the other hand the same Windows IMAP clients (we've tested both Outlook and Thunderbird) were far faster while connected to Windows IMAP servers (Exchange, MailEnable, hMailserver etc).
I am deeply fond of Scalix and therefore I just did not want to imagine their server was not-so-well-designed to allow so much difference against counterpart products for the windows platform (regardless the fact they're commercial - MailEnable - or Open products - hMailserver). More ... setting up an IMAP client on another linux machine to connect to Scalix (Kmail) offered reasonable performance far beyond the ones offered by an Windows IMAP client connected to Scalix.
So I had to look elsewhere for the right answers. And finally they came.
We were setting up a Samba server on our internal staging Scalix server: file transfers between Linux Samba and Windows clients (regardless they were clients or servers) was dramatically slow: on a 100Mb/s lan with only 2 computers connected on a dedicated switch we could not o beyond 10Mb/s nominal speed.
Googleing around to find any reasonable solution to this strange behavior I picked up this intersting page where is exactly described the reason and the workaround.
http://www.dd.iij4u.or.jp/~okuyamak/Documents/tuning.english.html
In other words ... TCP/IP layer used by linux machine is RFC compliant while Windows Winsock2 (it's own TCP/IP layer) somehow, is not and this leads to errors, doubled ACKs and dropped packets in communications between the two ends. Every flaw is a lack of performance.
I'm trying to get well documented about how to properly set up TCP/IP parameters on both the Linux and Windows side in a network environment where the two systems have to live toghether ... but ... is there anyone here been ever concerned about this problems ?
Any help or lead in the right direction is greatly appreciated.