January 11, 2008
Holidays delay
I must admit that I was a bit delayed by the holidays. First a one week LAN-party and then I got some sort of “winter” influenza, nothing serious but very bothersome. To top that off I still have to finish both my payed work assignment and school assignment in the coming days so in all honesty I didn’t work on lNet all that much. I did some minor changes, prepared packages for 0.6 and did some early fixes on the higher protocols after the send change. What still remains is the Qt4 visual support and at least some sort of SSL work (doesn’t mean it will be included right in 0.6.0 tho).
I’m not sure when 0.6 will hit the streets since my finals start 18. January and continue until 31. unless I fail some, in which case it would go on even longer. I hope to have released lNet shortly after this circus ends 🙂
November 6, 2007
0.6 incomming
I have started work on the first fundamental changes towards 0.6 release. Namely I have implemented the “easy sending for TCP” which enables users to just send data and not care anymore (more or less). Up until now you had to do a tedious loop with OnCanSend sending “trick” and in all honesty it wasn’t exactly optimal (usability wise, speed doesn’t change).
Now I need to test this on basic TCP server/client and start fixing higher protocols which use TCP. Old functionality will be still possible if you specify the old TLSocket as the SocketClass property for the TCP object. This means that even inherited TLSockets will work “old style” and of course analogically, new TLSocketTCP will also allow descendants.
Once this is done and FTP and HTTP works reliably again, I’ll start to hack on SSL. If everything goes well I might even manage a QT and Carbon Lazarus integration (will require probably Lazarus 0.9.26 or even greater depending on when my changes get in).
All in all 0.6 looks pretty promising.
On a side note I have entered benchmarking of the various event models (select, epoll etc.) as my research project at the university so we might get some tangible comparisons soon 🙂
September 16, 2007
SMTP bugs
Trom has came into #lnet recenently and told me of some odd problems with HTTP and SMTP he had. The HTTP problem is most probably a GTK bug in TMemo, but the SMTP problem turned out to be a bug in lNet regarding RFC compliance.
I had 2 problems.
1. The DATA command is defined by RFC to be “DATA <CRLF>” and one should wait for reply from the server before commencing the send of actual data. I had not only omitted the wait, but also used the format “DATA <first line of data><CRLF>”. Oddly enough some servers have accepted this happily. This has been fixed in 0.5.8 release.
2. The line length I used in SMTP was 74. This is a very outdated number, and since a long time ago, 1000 chars per line are allowed. There was also a bug in the line breaking algorithm which could cause data corruption on bigger attachments. I fixed both of these issues in 0.5.8 release.
I must say that I was honestly skeptical about Trom’s reports because SMTP worked reliably for most people until now. It just proves that skepticism isn’t always the best course of action.
Hopefully 0.5.8 is the last of 0.5 line and 0.6 will be released soon. More on that later…
August 12, 2007
Blog start
I have just been inspired by Vincent’s blog to start a blog of mine own so here it is.
lNet is progressing nicely. Version 0.5.8 will hopefully be out soon and I hope it will be the last of the 0.5.x line. There are some plans for simplifying of the sending in TCP sockets which will slightly change semantics and warrant a 0.6 release.
If everything goes ok the next major release will also contain OpenSSL support and possibly also some additional protocol implementation.
lNet - Lightweight networking library 