h2g2 The Hitchhiker's Guide to the Galaxy: Earth Edition

I don't entirely agree - it may be required less frequently, but the majority of fragmentation is generated by multi-tasking applications trying to write to the disk simultaneously, and "taking turns" to write a "packet" of data each... As RISCOS is multi-tasking, It can still suffer from significant fragmentation.

I've got a few Acorn machines lying around the house, from an "Atom" (circa 1980), through a BBC "B" and "Master Series" to an "Archie" (Archimedes A3000). They're nice to look at, but I can emulate any one of them faster than the real thing on my PC, so there's little incentive to crawl around behind the TV fiddling with RF leads...

Well, RISC OS is co-operatively multi-tasking, so it's not as bad as, say NT for fragmentation.

Point taken, though. Although 99% of Acorn users will never need to use a disc defragmenter throughout the life of their machine, and performance won't deteriorate noticeably.

Althought technically true, here is a fact: there are /no/ utilitys to de-fragment hard discs on RISC OS, and the only one to check how fragmented it is is part of the OS (ie, no-one else needs to write one) -- they are simply not needed.
All inter-application data transfers are via memory, and all apps take it in terns to write to discs.

-OPD

Up until a couple of years ago there were /no/ utilities to drfragment NTFS partitions. This can't be taken as an indication they were never needed. If you can't defragment a partition to compare, how can you know there will be no performance change?

Apps can't take it in turns to write to disk, in the way you suggest, else when capturing streams of video or audio nothing else would be able to write...

Uh... Are you *sure* that's the reason for other machines fragmenting?

I thought the reason PCs historically fragmented terribly on DOS was that it simply used an algorithm to select where to write on the disk which could be described more or less as 'fragment if at all possible'. Moreover there is no way in which MS-DOS could possibly be described as multitasking, but it seemed to have absolutely no trouble fragmenting the disk chronically.

By contrast the file system on RISCOS not only took pre-emptive steps to avoid fragmentation in the first place and IIRC it then made efforts to reduce it on the occasions it did occur (normally only when the disk got really close to being full).

It's the reason for multitasking OS's suffering from proportionately more fragmentation. DOS fragmentation was commonly caused by open files such as activity logs being appended to while other application files were also open. The first free cluster would be allocated to whichever file was written to next, leading to the files becoming "interwoven". It wasn't as bad a problem as with Windows etc., but DOS machines tended to be so low powered that any performance reduction was painfully noticeable.

Well, the original statement (i.e. RISC OS does not need defragmentation) is plainly wrong. RISC OS does need defragmentation, but the filing system does detect this by itself and does defragmentation automatically. As this is done (more or less) regularly and only locally it doesn't take much time (only a few seconds), thus most users don't even notice that it happens.

Furthermore the filing system always tries to prevent fragmentation as it tries to store the data at a podition on the disc, where the whole data can be written. This does work best (but not only) if the filing system does know how much data is to be written when creating the new file (of course). This means that defragmentation is needed much less frequently than with (most) other filing systems.

With the deepest respect, I suggest that at least one person on this thread gets a clue of some kind.

Disk fragmentation is largely caused over a period of time by files being deleted and then new files being written to the disk (much like dynamic heap fragmentation). I've never heard of multi-tasking being cited as being a reason for fragmentation occurring, and indeed it sounds like the sort of tosh Acorn owners used to make up to apologise for various features of the platform.

I mean, honestly!

As for RISC OS not needing defragging - well, I could never work out why if the RISC OS disc format ('E' format I'm referring to) was so great and always optimal, it was still way slower than disk access on my similarly specced PC.

Tim

PS. Before attempting to flame me, bear in mind that I'm someone who knows more about RISC OS disc formats than is healthy.

Re: "the sort of tosh Acorn owners used to make up", I posted the multi-tasking explaination as one of many reasons for fragmentation buildup. Bear in mind I'm someone who has used more multitasking OS's than is healthy...

I'm so glad someone said that - I was trying hard to avoid reverting to my bad old usenet habits...

As for why RISC OS disk access is slower than PC disk access on a comparable system, I think there are 2 areas in which PCs do far better; the relatively good fragmentation avoidance (relative to plain old DOS) merely mitigates against the other disadvantages: low-performance IO model (no async IO, no DMA on most systems) and less than brilliant caching.

One other thing: when was the last time you used a PC without any disk caching? Very occasionally I end up running in DOS with no frills, and it's painfully slow. Of course if you're used to running something like NT or Linux it's dead easy to forget that they can be quite that bad, whereas a lot of Acorn fanatics liked to pretend that no real advances were made after DOS 3...

> I'm so glad someone said that - I was trying
> hard to avoid reverting to my bad old usenet habits...

Me too - but Acorn enthusiasts plus fuzzy logic always makes me see red

I'll get over it one day.

Tim