Dear Fellow Open-source users:
Long before I was aware of open source, I had seen Word do much the same think on occasion. When I saw this happen, I knew I had to work very defensively and copy the saved file to an independent location every few minutes. I have seen this problem occasionally from Oo also, but frankly, I suspect it is a memory management issue. Also, if you are short of RAM in relation to your needs at that moment, the shortage in any system is going to end up in the swap file. This is slow, inefficient and lacks stability. Writing to and reading from a hard drive is not as simple as it might look, and there are always many possibilities for error. The more the swapping the more the chances for trouble, be it this way or in the process of the OP Sys sorting/defragging the swap file on the fly. When I was running OS/2 (Warp) on my custom made machine with the double SCSI chain, I had a physically separate HD just for the swap file, and I manually edited the config.sys file to redirect the swap. At the time a 1 Gb. SCSI 2 drive was ok for that. Added to that, the concurrent seek timing benefits of the physically separate drive would have helped somewhat also, depending on the SCSI chipset, the motherboard processor and the onboard chipset management of the PCI bus on which the SCSI would have sat. Soon, I plan to build another special machine, and run Fedora on it. With the help of my Linux guru friend (who holds a masters' degree in IT), we expect to redirect the swap file to a separate drive. At the time, I would like to use an SSD for the purpose as it would eliminate some of the seek and mechanical issues as well as hopefully get rid of some of the Encoding/decoding burden needed with magnetic media as well as reduce heat and power consumption a bit. This will depend on the cost per Gb. Trend in the market at the time we do this. If that is too high, I can easily get a 1 Gb. SCSI 2 drive for $C10.00 and use that until the market cools off. Cables and stuff are no problem for me. I have made my own internal SCSI cables and a termination in the past, and still have some of my old stuff around. (The SCSI cable for the past special machine was 2 M long and had about 15 connectors for various options.) I suspect redirecting the swap in Win might be a touch more difficult, but if one is going to put up a multiple OS system, swap redirection is a separate matter for each OS, and would likely be better not to share the same swap drive. At $10 a pop, it would be better to use a second drive for Win if you really wanted to bother, but you would also want to know the max RAM Win 7 would take and have a motherboard and RAM to match. If the SSD swap does well on a machine that is tested with a deliberate shortage of RAM, that might also provide a partial hedge against the aggressiveness of the market in regard to the Motherboard/Processor/RAM combination. Also, I would like to see a decent motherboard out with a socketed BIOS chip, so if one has a flashing failure, the BIOS chip can be easily replaced (or, to be safe, take the BIOS chip and have a copy made on a PROM burner before flashing onto a virgin chip.) (After all, I really wouldn't want to have to go to the trouble of taking a hot air rework station to get an SMT chip off the board and replace it with a socket and Electrically equivalent chip that was socketed! That is not so easy, having done some of this kind of SMT component level rework in my time.) Another possibility would be to have a socketed EEPROM driving an FPGA and an onboard socketed SSD for a BIOS that was a bit more than just basic. All in all, hybridism is more complex, but with that added complexity comes added versatility and often performance which comes from the uniqueness of each of the hybridised elements. (I wonder if this will make Tux smile!) Cheers, Bruce M.
