[Raion]

I was clearing out my old email backups today when I saw, curiously, a specific thread I've been wanting to share for years. 

One of my many attempts at getting ahold of someone at Rackable who would release the IRIX source or talk to someone, Gentoo-MIPS dev Kumba asked for me to check if they had any documentation lying around.

Here's the inquiry after I was connected to this individual, and the ensuing conversation we had regarding both IRIX source code, and the documentation story:

To: Kazuo K.<b****fox@t***********.com>
From: Bob <b***@sgi.com>
Date: March-30th-2015 9:32 AM


Hello Kazuo,

Unfortunately I doubt we have anything of that vintage here. Back in 2010, a year into my working here at SGI, several staff were called into then CTO Goh's office for a request. Goh was eager to get piles of documents shredded that dated back to the mid 1980s and had been inherited from the old SGI. He advised the staff to carry this out quickly. As such, several hundred pounds of documents of 68k and MIPS era stuff were simply destroyed over a three or four day period. 

It's not impossible that there's still copies of these out there, but you'd have to track down old engineers for that sort of answer for paper stuff. For digital, I'll continue asking around for you and see what we can do.

In regards to your request for IRIX source for things like MIPSPro, that's probably never gonna happen from us because parts of it are licensed by Edison Design Group, and that's going to be a massive issue. IRIX itself has so much code from Sun, Microsoft, AT&T and the old BSD in places that it may be a nightmare. XFS is unlikely to be released, but I can see about XSGI and Indigo Desktop. 

Thanks,

Bob [REDACTED]

---------------------------

To: Bob <b***@sgi.com>
From: Kazuo K.<b****fox@t***********.com>
Date: March-30th-2015 11:29 AM

Hey there Bob. Thank you for checking on that. I'm really kind of appalled that they'd wash their hands of that... so much information just gone through a shredder. Yes, please check for a digital copy or scans. Heck, I'll even take a microfilm!


I've asked since like 2014, so I'm not surprised the answer hasn't changed. But I guess I won't give up? IRIX being opened up is a long term goal. 

Best Regards,

Kazuo K. 

---------------------------

To: Kazuo K.<b****fox@t***********.com>
From: Bob <b***@sgi.com>
Date: April-12-2015 3:10 PM


Hello Kazuo,

I was unsuccessful at attempting to track down a digital archive of any of the engineering documentation on those machines. It's possible it's still here, but it's been lost through staff turnovers. I'm not very optimistic. I would try tracking down engineers from that vintage of the company. I bet some of them have copies of their manuals at home they could be coaxed to scan or let you borrow. 

I'm sorry I couldn't find anymore on that. I appreciate you understanding that. 

Best,

Bob [REDACTED]



At some point I passed bob's email to Kumba I think? But things never got anywhere. It's unfortunate.

Anyways, that's that really. I considered sharing full headers and such, but I really don't want to expose my domain name or other PII about me or Bob. I've already got people who are trying to call me by my roommate's name or other mistaken identities.

Hope that's useful to you guys here for citation.

[TruHobbyist]

Thanks for sharing. I enjoyed it. What I keep asking myself is why all these proprietary UNIXes share so much code among each other when it is really simple to provide clean room implementations of technologies (take VFS as an example) or just create your own version of it (let's say Microsofts/IBMs IFS vs. VFS).

Maybe someone could shed some light onto this.


As per IRIX, I think it is not an absolute requirement to get hands on their internal documentation. With the IRIX source code floating around you have everything you need to know. Software and hardware are like a plug and its corresponding socket: it's enough if you have one of them to *functionally* recreate the other. But of course, it would be a nice-to-have thingy.

[weblacky]

What I keep asking myself is why all these proprietary UNIXes share so much code among each other when it is really simple to provide clean room implementations of technologies (take VFS as an example) or just create your own version of it (let's say Microsofts/IBMs IFS vs. VFS).

Maybe someone could shed some light onto this.

I was always lead to believe that (back then) you cannot be a UNIX without being derivative work (now I thinks that's not true...you can receive UNIX compliance and certifications).  But originally, wasn't it the case that to be a UNIX you had to share source from the AT&T kernel and have a certain amount of POSIX calls to be implemented for a UNIX OS?  Otherwise you're a clone?

[jan-jaap]

He claimed in 2015 that XFS was unlikely to be released? Interesting.

[Raion]

I was asking specifically about the IRIX XFS implementation.

[vishnu]

I was asking specifically about the IRIX XFS implementation.

So wait, you're saying that the MIPS specific source code was ripped out of XFS before SGI open sourced it? I don't use XFS so I haven't looked, but I don't think they did. Or, perhaps more to the point, if they did it was a totally craptastic thing to do.

[Raion]

The released source of the most early versions has an emulation layer to interface to the Linux kernel filesystem. SGI made a system called Vnode for IRIX, which is a specialized VFS implementation. It's broadly similar to some BSD ones, but it's different in important ways. Later versions of XFS AFAICT have ditched this

I haven't trawled through the code in ages, and because I want to avoid GPL code when possible (I.e. when it's unnecessary for me to do so, unlike say, having to fix bad GCC code) it's unlikely I'm gonna do it anytime soon.