SGI: Development

Actually, transferring pretty large files via email is possible. Just usb mass storage support would be several degress of awesome.

zizban wrote: Actually, transferring pretty large files via email is possible. Just usb mass storage support would be several degress of awesome.

Yep, the ability to slap a couple of terabytes of USB storage with relatively low power requirements onto an IRIX box for less than $100 without having to worry about glitches or gotchas would be terrific. Unfortunately, I doubt that we're going to see that happen.

As an aside, I use a 10 Mbit/s Challenge S as a bridge between my two home nets. Works fine, and it still has more bandwidth than my backup DSL connection to the internet. That said, I probably will update to a faster bridge, since my cable modem connection to the internet is nominally 20 Mbit/s. I'll keep the Challenge around as a print server, however. Just because...

If you have someone with the skills to implement USB block devices on IRIX then they shouldn't have a problem with an AHCI driver, right? Then you just pop one in the PCI or shoehorn. SOL if you have an Indy/Indigo2 or earlier, but...

Granted, I'm so glib because I know I have not the skills to do either, and won't have them anytime soon, for any definition of soon up to and including geologic time. I'm firmly out in userland as far as programming goes, and don't even do much to speak of there.

While we're on the subject of USB drivers - hamei just mentioned to me that he could probably hire someone <with the skills necessary to do USB driver dev work and other programming issues> in China for a small fraction of what it would cost on this side of the great wall. If anyone is interested in getting involved PM me your contact info and I'll pass it along.

SAQ wrote: Granted, I'm so glib because I know I have not the skills to do either, and won't have them anytime soon, for any definition of soon up to and including geologic time. I'm firmly out in userland as far as programming goes, and don't even do much to speak of there.

...and I guess that's the bottom line. While most or all of us would love to have access to full USB compatibility (or Firewire, too, lol), there are few, if any, of us who have the necessary combination of time, skill, motivation, and knowledge of IRIX internals to make it happen. I wish it weren't the case.

If someone does manage to do it, they will be praised to the hilt, but that's pretty much all that they are likely to get for developing a USB driver for a platform that last saw a newly manufactured machine four years ago and that will be unsupported in two years. Anyone with the requisite skills to do it would be able to do something much more profitable with their time, so it would have to be a real labor of love.

recondas wrote: While we're on the subject of USB drivers - hamei just mentioned to me that he could probably hire someone <with the skills necessary to do USB driver dev work and other programming issues> in China for a small fraction of what it would cost on this side of the great wall. If anyone is interested in getting involved in that project drop hamei an email <or PM me your contact info and I'll pass it along>.

It's an interesting idea, but even with Great Wall labor rates, it wouldn't be trivial, particularly if real QC/QA, i.e. reliability, is needed. Still, if hamei thinks that can be accomplished with a reasonable investment, and he uses his considerable charm to ensure QC, it might be worth pursuing.

Any idea how many yuan he thinks would be needed? Is it within the realm of a small community contributing shareware money, or is it something requiring some significant philanthropy?

Definitely more long the lines of a small community shareware project - primarily among those who know hamei well enough to be comfortable with the arrangement. I think he wanted to see what <if any> interest there was before proceeding.

A safe way to do it with minimal coordination effort might be through Kickstarter - someone proposes a fundraising project with a budget, people make pledges to the project via Amazon checkout, but no money changes hands unless the project is fully funded. If the project doesn't get fully funded, no one gets charged anything. I'd be willing to pony up some cash.

A musical acquaintance of mine used the mechanism to fund the manufacture and distribution of her latest cd - pretty neat, actually.

I see just one big trouble on that: Where to connect the USB adapter. I think THIS is the big problem. I have no fast enough interface on my octane to connect it without troubles.

Tabalabs wrote: I see just one big trouble on that: Where to connect the USB adapter. I think THIS is the big problem. I have no fast enough interface on my octane to connect it without troubles.

You'd need to find an SGI "PCI Card Cage" or a PCI to XIO adapter (also known as a "shoehorn") and then add a PCI USB card, assuming USB drivers could be developed. See viewtopic.php?f=3&t=16720341 for ideas about other PCI cards that work in an Octane.

The hardest part is the driver, which does not yet exist. The hardware to use PCI boards in Octanes does exist, but may not be easy to find in Brazil.

josehill wrote: A safe way to do it with minimal coordination effort might be through Kickstarter - someone proposes a fundraising project with a budget, people make pledges to the project via Amazon checkout, but no money changes hands unless the project is fully funded. If the project doesn't get fully funded, no one gets charged anything. I'd be willing to pony up some cash.

A musical acquaintance of mine used the mechanism to fund the manufacture and distribution of her latest cd - pretty neat, actually.

I was thinking Kickstarter as well.

Oh yes, I'd die for a PCI card cage...
But since there IS hardware for that, I don't think the hardware is too far away. Remember: There is a complete and working driver for linux, and it cannot be SO different

Tabalabs wrote: Oh yes, I'd die for a PCI card cage...
But since there IS hardware for that, I don't think the hardware is too far away. Remember: There is a complete and working driver for linux, and it cannot be SO different

You've never written a kernel level device driver have you?

I did the Solaris version of a SCSI device driver for large medium changers.. Massive media libraries driven by crypts, drives and a robotic arm for government, big corporates and the like.. I also did maintenance for the OSF/1, HP-UX, early AIX, and WindowsNT drivers... the code all served basically the same function and was a million miles away from each other.. I also did a Linux device driver, while I was a post grad student.. Read the "Magic garden explained" and "Understanding the Linux Kernel" and tell me if they are the same book.

Actually I have been working on this and got bored with it and stopped.
I bought 3 books on USB.
It would require 3 device drivers...
I started a vt6212 driver for a PCI card I put in the Octane. Then you would need another driver for the other USB device class that the IRIX doesn't support ... the slightly more complex one, not OHCI the other one... and handle the USB request blocks for it (URBs)..
Finally.. you would need a upper kernel block/filesystem/class mass storage driver.
After all that you would need userland programs like "usbprobe" or something.
I was going to shut the fuck up about this until I had something I could prove did something at all, even from userspace. Now I will wait for everyone to get bored with the topic again.


R.

I have started a Kickstarter account for this project. Here are the questions that need to be answered. I have put down suggested answers we can tweak before I submit.

What would you like to do?

Create a userland usb mass storage drive for Irix. Irix is an Unix operating system with solid hardware to run on. However, Irix is no longer being developed and the community would like to extend it's functionality. The goal is to raise money to develop an open source USB mass storage driver to extend the usefulness of of Irix.

Money raised will be used to acquire equipment and to offer bounties to developers.

What rewards would you offer?

All contributors will receive a T-shirt or a USB flashdrive with the project's logo.

Where can we find out more about you and your project?

www.nekochan.net

The goal will be to raise a $1,000-$2,000

Chime in with suggestions.

zizban wrote: Chime in with suggestions.

I've been paid for cutting code since 1986.

The best way to a project failure is a all or nothing approach on a grandiose and overly ambitious plan. Start small have a small goal then build on it.

Small project, quick and trivial outcome, slightly more involved project based on the foundations of the previous effort and build from there, I've had teams who worked for me deliver stuff on time and on budget and actually praise my leadership publicly.. In one case they had a success after two days, another after a week and then major milestone delivered to user acceptance testing after a little more than a month. Where as I've worked on projects that blew up with panicked project managers throwing everything at it and just hoping the whole thing works with no milestones to measure success or red-flag problems.

Its drinking season here now, with vendors like Sybase, Oracle, etc throwing end of year parties..
And I 've a new Tezro to play with.


I suggest break the project down into much smaller and manageable pieces.

3 drivers are required and even then, that can be broken down into much smaller pieces..


R.

zizban wrote: The goal will be to raise a $1,000-$2,000

It should not be difficult at all if it awakes some interest between the forum members. Count me in to contribute some money if the project advances. This is one of the more interesting additions we could count for IRIX.

PymbleSoftware wrote: The best way to a project failure is a all or nothing approach on a grandiose and overly ambitious plan. Start small have a small goal then build on it.

Regan's right. If people are serious, someone who knows what they are doing from a development perspective should identify a small, but meaningful piece of work, make that the first project, demonstrate success, and then proceed from there.

josehill wrote:

PymbleSoftware wrote: The best way to a project failure is a all or nothing approach on a grandiose and overly ambitious plan. Start small have a small goal then build on it.

Regan's right. If people are serious, someone who knows what they are doing from a development perspective should identify a small, but meaningful piece of work, make that the first project, demonstrate success, and then proceed from there.

I agree with you. A small sample should rise the expectations a lot, and will probe the available firepower to the more hesitant. Mass storage is one of the biggest problems when you can't have good access to big/fast SCSI drives, and let's be honest: SCSI is everyday older, and the bigger sized drives are not so common for some countries... and the second hand market vary a lot.

Would not be fantastic to be able to use the same widely available drives from PeeCees?

IRIX USB support X-Prize? Count me in for $50.

GeneratriX wrote:

josehill wrote:

PymbleSoftware wrote: The best way to a project failure is a all or nothing approach on a grandiose and overly ambitious plan. Start small have a small goal then build on it.

Regan's right. If people are serious, someone who knows what they are doing from a development perspective should identify a small, but meaningful piece of work, make that the first project, demonstrate success, and then proceed from there.

I agree with you. A small sample should rise the expectations a lot, and will probe the available firepower to the more hesitant. Mass storage is one of the biggest problems when you can't have good access to big/fast SCSI drives, and let's be honest: SCSI is everyday older, and the bigger sized drives are not so common for some countries... and the second hand market vary a lot.

Would not be fantastic to be able to use the same widely available drives from PeeCees?

What do you suggest should be the first small step then?