Well, I must admit I haven't used Calmira much for the last few months,=20 since I've been quite busy and had enough boxes to play with that do not = need Calmira. Recently I dug out my 486 though, finally installed the=20 Promise EIDE2300+ VL IDE controller (this allows me to max out the hard=20 drives), messed up Win95 when installing the driver not according to the = readme, couldn't get it reinstalled without crashes, put NT 3.51 on E:=20 (the second partition of the 1 gig drive) - that worked just fine -,=20 finally got Win95 reinstalled and working halfway properly (truetype=20 fonts do not work properly for some obscure reason, and Scandisk says my = partitions may not have been formatted properly), then discovered=20 extensive file system corruption on E: (darn, shouldn't have put the=20 temporary install file for 95 there *grmbl*), lost NT because of this=20 and frustratedly put the old box back into its corner. A few days ago I=20 decided to give the old beast another chance - the last one -, ripped=20 out the Pentium Overdrive and put the DX/2-66 back in, reinstalled NT=20 and set up the programs needed, defragged all the partitions in 95 and=20 have been quite happy since, except for the original driver for the=20 graphics card refusing to install and the performance when running=20 Mozilla (well, and I can only play back MP3s at 22 kHz mono in NT, while = IIRC in WfW 22 kHz stereo worked, too). Normally its rendering should be = faster than Netscape 4's (it least that's the case on my more modern=20 boxes), but in this case this doesn't apply - I guess that's because of=20 the abysmal memory performance of the OPTi 495SX chipset: DXOVER's=20 System Speed Test 4.75 measures 9 MB/s for reads from memory (and that's = with all kind of tweaks), while a board based on the 895 manages 35 MB/s = at 40 instead of 33 MHz, and is 50-100% faster elsewhere. I guess the=20 fact that Netscape 4 performs better than Mozilla in terms of rendering=20 is due to it using smaller data structures that fit into the cache=20 easily, while Gecko uses more memory and possibly accesses it more=20 frequently. (ISA performance isn't too great either: When clocking the=20 bus at 5.55 MHz, the machine doesn't manage to properly play back a=20 simple 44.1 kHz 16 bit stereo WAV file residing on a network share, that = should be a data rate of 2x 172 KB/s + x.) I guess I'll be getting a=20 board based on the newer OPTi chipset (with PS/2 memory slots, but I=20 could get hold of adapters to PS/2 rather inexpensively), which would=20 not only give better performance, but also offer a 3rd VLB slot (out of=20 spec at 33 MHz, but maybe it works anyway - it'd be interesting to=20 research the CPU usage of busmastering VL SCSI host adapters) and maybe=20 solve the graphics driver problem. BTW working with old Program Mangler and Task Manager is pretty clumsy - = Calmira would make a nice shell for NT 3.51 *if* it'd be possible to get = hold of NT's long file names somehow. I imagine one would have to write=20 some 16-to-32-bit interface DLL for that (and possibly do a Windows=20 version check on startup), not the easiest task but hopefully not=20 impossible. (At least that's more realistic than Calmira being ported to = Win32 - Delphi never was an NT 3.5x thing.) As of now, Calmira works in=20 NT 3.51, even if a tad slow. Stephan --=20 Stephan Gro=DFkla=DF (7bit: Grossklass) an der Hauptkiste: P3B-F,P3-500@620,512MB,20+18+8GB HDD; Win2k, WfW 3.11 + Calmira II 3.2 eMail: mailto:jgrossklass@xxxxxxxxxxx | Webmaster: http://www.i24.com/ Home: http://jgrossklass.bei.t-online.de/ -- To unsubscribe, send a message to ecartis@xxxxxxxxxxxxx with "unsubscribe calmira_tips" in the body. OR visit //freelists.org