Gary Kildall invented CPM for the Intel 4004(?) which sort of launched a new market.
Never sold as a 4004 CP, only on 8 bit micros and bigger
went on to found Digital Research which created GEM. GEM got the pants sued off of them by Apple (there is a surprise) which slowed GEM's development a great deal.
Yes and no, GEM actually predates Apple developments, it has its origins in the GSX graphics library that was very popular for engineering and scientific applications but almost unused in home computers except for one Amstrad model, I had a working multi processor CP/M system with a rudimentary GSX based user interface until 2000 when I was forced to throw it away due to a move, that system had not been upgraded since 1982 so it predated the Lisa.
DR did not lose the Apple lawsuit but decided to agree to change some things that were Apple like
However, when GEM was licensed to Atari it wasn't as restricted at it was before but you can see that GEM was pretty ugly all the way 'round.
GEM was not ugly, it was like the early versions of the CP/M 68K optimized for the PC, and the best selling graphics interface on the PC was a CGA so everything was optimised so it would look OK on 320x200x16 which was the only CGA mode that could be used bitmapped, later versions like GEM/3 actually looked better than the Windows versions that shipped at the time.
TOS was a disaster for Atari and you will know why if you have read my other rambling posts here. TOS was designed on Apple Lisa Computers but TOS uses instructions that are not available on M chips > than 68k (010,020, etc...).
As I have rambled on in other posts, I cannot conceive of why you would do this. More to the point, why you wouldn't fix it in a revision of TOS. TOS 1.0 (like AmigaDOS 1.0) was a piece of crap so rewriting TOS to use certified instructions might have broken compatibility with TOS 1.0, but who cares when you have such a major problem? Instead, they just kept rolling out TOS versions that only cemented software to the problem.
DR sold the rights to an early CPM68k port and GEM version to Atari, note that this version does not correspond to a release version of either system, this was meant to be a development version and a later version was meant to be used with the shipping unit. Remember that early versions of Z8000 and 68K CP/M ports very full of 8086isms some intended some not, this got better later on, but the original idea was also compatibility so that programs could be ported from 86 to 68K, then to Z8K and so on which explain some of the 86isms in the DR products
DR wanted payment for the final version of both and Tramielsky said no, DR say bye and refused to release any of their programming languages or other software on the Atari platform even tough that was the original intention, this created panic when the original ST dev systems were released with useless development software.
Atari also later added FastTOS routines to TOS and actively encouraged programmers to use them, this killed compatibly with other GEM systems but there were a few software houses that never used the FastTOS so they could easily sell DOS versions as well (Artline for example).
Atari had to rewrite TOS with the TT (the graphics workstation without a blitter) and the Falcon and this was a major problem with backward compatibility. Commodore had a similar issue with AmigaOS 2.x since it broke poorly written Amiga 1.x programs. I suspect that TOS and Muti-TOS was a much harsher upgrade given the extent of the rewrite.
While Multi-TOS was a nice upgrade, it just added to Atari's growing list of problems.
Multi-TOS was stupid, especially the decision to keep it 68000 compatible, if they had focused on 68030 it would have made sense
BTW: Gary Kildall was on Computer Chronicles and is dead (I believe some sort of bar fight or accident?)
And sadly missed, died of a injuries sustained during a drunken brawl, funny really because he was viewed as one of the nicest blokes in the computer industry....