This has been builtin to Bitcoin-Qt since 0.6 and used for things like the Eligius mining pool and #Bitcoin-OTC for just about as long.
Here's an example walkthrough written by BunnyH.

I don't get why people still tolerate cgminer and all its many bugs.

This is a cgminer bug. When the nonce2 is smaller than 32-bit and not followed by null bytes, it corrupts the coinbase transaction and thus produces invalid shares.
BFGMiner always uses a 16-bit nonce2 (plenty big enough) to ensure it can work with upstream nonce2s as small as 24-bit.

Edit: Details:
[04:23:17] <Luke-Jr> nwoolls: basically cgminer is broken if nonce2_size is anything other than 4 (32-bit)
[04:23:41] <Luke-Jr> nwoolls: a proxy must always reduce the nonce2_size from the upstream server; BFGMiner always uses 2 (16-bit)
[04:24:14] <nwoolls> Luke-Jr: can you pastebin the patch that makes it work so i can at least give that stuff to Con and say I tried?
[04:24:30] <nwoolls> i'll hit him up when he's around and see if he's feeling cooperative
[04:24:27] <Luke-Jr> nwoolls: I'd have to write one.
[04:24:31] <nwoolls> ah gotcha
[04:24:38] <Luke-Jr> I'm not sure it's worth the effort, considering the likelihood they reject it
[04:24:48] <nwoolls> ya thats ok, i just meant enough for me to convey it to him
[04:24:42] <nwoolls> so those 2 lines should be enough to understand the issue for him?
[04:25:02] <Luke-Jr> nwoolls: probably not, those 2 lines are sufficient to explain the issue generally in a way that he will probably outright deny.
[04:25:19] <nwoolls> lol
[04:25:24] <nwoolls> nice
[04:25:27] <nwoolls> well i'll give it a shot
[04:25:34] <nwoolls> maybe he'll deny it and then look into it himself
[04:25:32] <Luke-Jr> nwoolls: after you confirm that hypothesis, you can give him more info: he is memcpy-ing nonce2 with a fixed 32-bit size, even if nonce2_size is smaller
[04:25:52] <Luke-Jr> nwoolls: he is also intentionally rejecting nonce2_size over 4 (or 8?), IIRC
[04:26:07] <Luke-Jr> which is perfectly legal, but in any case unrelated to BFGMiner interop issues
[04:26:32] <nwoolls> ty for looking into it, i appreciate it, starting to hear it from more and more users
[04:26:41] <Luke-Jr> side note: if he just memcpys the nonce2 with an adjusted size, it will fix *little endian*, but break on big

Edit2: <rant about cgminer bugs still wasting his time; 2.5 hours in this case>

Difficulty does not affect devices at all.
Even if you could set difficulty, it is unlikely to allow setting it lower than the automatic variable difficulty (which is lower than most pools use on Eligius).

I've collected a few on my site: http://luke.dashjr.org/education/tonal/glyphs/fonts/

Difficulty changing is not harmful in any way.
Stale shares just come from latency on your internet connection and/or hardware that can't change work on a new block.

That sounds correct. It's very hard to get diff 1 with scrypt.

I don't really care about scrypt.
If you (or anyone else) can bisect it, debug it, or provide a patch, I'll look into it...
If you use Windows, you can easily bisect by starting a new session on http://luke.dashjr.org/tmp/code/webisect/webisect.php

No, it really needs to be rewritten differently (as expected).
Unfortunately, that's very involved.

I can confirm twinfury is developed by the same as red/blue fury.

Yes, a USB CDC ACM device, so would be /dev/ttyACMn form.

Felipeo said it would be on http://store.bitcoin.org.pl/support a few weeks ago, but I can't seem to find it either

No, this looks more like a broken USB hub or insufficient power...

The shipped firmware for some (most?) of the bi*fury is buggy, and an updated firmware is needed to get them to mine correctly.

It requires writing code.

FYI: First batch of HashBuster Micro sold out and the next batch is delayed slightly until Monday. Orders placed before then will ship direct from manufacturer.

Eligius is an entirely volunteer run pool.
If you'd like to write a new OP, I'll change it.

BFGMiner is probably broken on FreeBSD. Any FreeBSD developers who can help out?

Namecoin is bitcoin technology applied to an entirely different concept (domain names instead of currency).

Using getwork for something like this isn't going to work.
Check out README.ASIC for instructions to install BFGMiner directly on your KnCMiner unit.

Essentially it comes down to there being a void of any reason people would adopt Litecoin as a currency.

Bitcoin by nature can only provide the same value put into it, so in this "raw" state it functions as nothing more than a pump-and-dump scam as it simply redistributes money from the buyers to the sellers.
The reason Bitcoin itself escapes this category is that it is a major technological innovation, bringing something new (trustless currency) that has value to the world.
By being adopted as a currency, Bitcoin can benefit even those who buy into it last.

Litecoin and other scamcoins like it, on the other hand, do not bring anything new to the table.
They are just mere clones that retain the pump-and-dump nature of Bitcoin, but without the innovation that makes Bitcoin viable as a currency.
Litecoin specifically made three irrelevant changes:

• Change of mining proof-of-work from SHA256d to scrypt. This was done as an attempt to make GPU mining impractical. Not only is that a bad idea (CPU-only mining would put cryptocurrency in the hands of illegal botnets), it also failed. Present-day scammers will try to claim scrypt is ASIC-proof or ASIC-resistant; this is also false and impossible in theory. If anything, scrypt is more vulnerable to ASIC mining since it performs poorly on consumer hardware. Finally, some people try to pass scrypt off as a memory-hard proof-of-work; while it is true that the scrypt algorithm itself is a memory-hard key-strengthening algorithm, it is not a memory-hard proof-of-work algorithm since it requires just as much memory to verify as it does to find (the key property of proof-of-work algorithms is that the work requires more of <resource> than the verification); Litecoin's scrypt parameters also use so little memory, that it can be performed without memory at all.
• Faster target block time. This is often passed off as "faster confirmation", but in reality it isn't at all. To get the same security as 6 bitcoin blocks confirming a transaction, you need around 24 litecoin blocks (there are a lot of factors involved that weigh into this). In the end, the faster blocks just bloat the blockchain more - and not just the entirety of the blockchain, but the headers-only minimal blockchain that would be used by light clients.
• Larger currency supply. But we all know 1 BTC is divisible to 8 decimal places. There are more mBTC than LTC, so no "silver to gold" is needed in an alternate blockchain.