25 dollar computer coming

[curaga]

USB is just about the worst bus to swap over, though. When under memory pressure, hey let's add cpu pressure too

One can swap over network, but I don't think that's still 100% reliable.

[tinypoodle]

There are thousands of non-solidstate USB drives available.  How could you not find one?

Not to mention that they predate solidstate...   

[caminati]

There are thousands of non-solidstate USB drives available.  How could you not find one?

Gee... sorry, I forgot to mention that in my raving rant there is no space for magnetic devices: they have moving parts, they are still too big and cumbersome for the roving user.

Never liked that technology, though I must admit they managed to attain wonderful specs with it now. Also, you can fit 128 Gb on a microSD nowadays: you will never get such a high capacity/size ratio with those frantic magnetic disk rotators.

I was thinking about something like sdram or sram on usb. There are newer technologies which would be suitable (T-ram, Z-ram, etc...), all more or less in infancy though:

https://secure.wikimedia.org/wikipedia/en/wiki/Computer_memory

I am admittedly just daydreaming, anyway.

[danielibarnes]

The HyperOS HyperDrive5 (64GB max: $450) and ACARD ANS-9010 (64GB max: $400) are two examples that come to mind. I expect you could use eSATA or put them in a USB-SATA drive case.

[curaga]

Microdrives are CF sized, are they small enough?

[caminati]

The HyperOS HyperDrive5 (64GB max: $450) and ACARD ANS-9010 (64GB max: $400) are two examples that come to mind. I expect you could use eSATA or put them in a USB-SATA drive case.

I guess those are the closest implementations, gosh they're big!
Maybe serializing the interface they could get smaller: those units just pack some standard ddr (or the like) modules into some adaptor, if I get them right.

@Curaga: interesting, I had no competence to evaluate that point.
I understand that usb is one of the communication protocols relying the less on hardware and the most on software, is that arguable?
I know of a bunch of overlapping standards for data over a net: AoE, iSCSI, FCoE.
Sadly they all pass through an interface other than usb.
The wonderful page (in which I feel close in spirit with TC principles in many ways):
http://harmful.cat-v.org/software/
promotes the first as the least harmful option =)

[caminati]

Microdrives are CF sized, are they small enough?

Nice! But you are trying to break a stolid bias of mine against magnetic media: no way =)

[gerald_clark]

Not to mention that they are quite fragile.

[caminati]

Not to mention that they are quite fragile.

Yes, and more noisy, and sucking more power: I'd say it's all due to the fact that they try to realize via the macroscopic, that is mechanically (Winchesters were born decades ago) what can be done microscopically.
This brings all kinds of complications, which they managed to tame surprisingly good.
But I find sort of wasted all this mastery in facing what could just have been avoided =)

[caminati]

This Raspberry in particular made me ponder about how nice it would be if even ram could be extensible via USB: one poorman's way would be creating linux swap file/partition on a usb flash drive, but then there are at least two issues: speed and write cycles wear.

Ok, I didn't know that many people already built less-poorman ways of exploiting more and more available flash memories. Two of the most interesting are Microsuck's Windows 7 Readyboost and DragonFly's swapcache. Here are some information links about them and some quotes relevant for the two issues cited above.

http://en.wikipedia.org/wiki/ReadyBoost

Since flash drives wear out after a finite (though very large) number of writes, ReadyBoost could eventually wear out the drive it uses—though this may take a long time, depending on various factors. According to Microsoft, the drive should be able to operate for at least ten years.[1] As capacities rise and cost per megabyte drops, USB drives are increasingly suitable for ReadyBoost.

and

http://leaf.dragonflybsd.org/cgi/web-man/?command=swapcache&section=ANY

The algorithms the SSD implements in its firmware are probably the most important part of the device and a major differentiator between e.g. SATA and USB-based SSDs.  SATA form factor drives will universally be far superior to USB storage sticks.  SSDs can also have wildly different wearout rates and wildly different performance curves over time.  For example the performance of a SSD which does not implement write-decombining can seriously degrade over time as its lookup tables become severely fragmented.

The bottom line could be that the very rough idea is to take advantage of the technical properties of solid state memory (low seek time, hence better for small scattered cache operations than magnetic media) rather than dumbly place a generic swap file on the device.

Also, usb is slower than regular sata SSDs, which doesn't mean, given the laughable current prices of usb pen drives, that the first option cannot help.

[remus]

It might be good hardware for a printer server, would not take up much space thats for sure.

And they do make usb/parallel adapters, great for some of our older printers.

http://www.shopbot.com.au/pp-belkin-usb-parallel-printer-adapter-f5u002-price-106075.html

[caminati]

It might be good hardware for a printer server, would not take up much space thats for sure.

And they do make usb/parallel adapters, great for some of our older printers.

In that case, since you most likely would have to add a usb hub, this thing could be suitable, killing two birds with one stone:

http://www.amazon.com/D-Link-DSB-H3SP-3-Port-1-Parallel-1-Serial/dp/B000023VUJ

The only drawback is it features usb 1.1, not 2.0 apparently.

[Rich]

The only drawback is it features usb 1.1, not 2.0 apparently.

It's plenty fast for a serial port and should be fast enough for most printers in text mode.
The drawback I see is that 3 out of the 4 reviews where less than favorable.

[remus]

The only drawback is it features usb 1.1, not 2.0 apparently.

Great Idea caminati, pitty that one is usb 1.1, i've done a bit of googling but have not found anything better so far, will keep you posted.

[caminati]

The only drawback is it features usb 1.1, not 2.0 apparently.

It's plenty fast for a serial port and should be fast enough for most printers in text mode.
The drawback I see is that 3 out of the 4 reviews where less than favorable.

Keep in mind that this yet-to-come raspberry would have only one usb port, so a hub would be a pivotal component to which not only slow devices would likely get attached.