Skip to contentSkip to conversation

UASP: USB Attached SCSI (UAS) protocol driver

PATCH Luben Tuikov 6 participants 10 messages Dec 10, 2010 – Dec 14, 2010

lore
Share & tools
lore
#0 lore raw

Before I begin, I'd like everyone to note that Greg has completely ignored (and removed from the quotation) the 18 points of technical matter in my email and instead decided to write this, to which I'm responding here:

Quoted 6 lines · Greg KH

What I don't want in _anyones_ kernel is multiple drivers wanting to
attach to the same device.

By "same thing" I mean "binds to the same device."

Thanks for clarifying, Greg. For a moment everyone thought that by "the exact same thing" you meant "same functionality, same quality, identical, etc".

Now, if you don't want the still-born uas.c to bind to the same device, please simply remove the ids from that driver.

Quoted 3 lines

I have lived through the hell of this in the past and I will not let
that happen again, sorry.

So, according to your statement above "binding to the same device" is all it takes. The solution is very simple--see above.

So to expose the obvious, hypothetically, one of the resident "linux kernel engineers/maintainers" could've submitted AN EMPTY DRIVER that "binds to" the UAS id, and from then on, any other improvements would have to go through them, even if they don't work with the technology and are not very (at all?) knowledgeable about the technologies/standards/methodologies/termninology involved.

Quoted 1 line

I'm not saying that the in-kernel driver is perfect at all,

I think you did say "the exact same thing" ;-) in your previous email in this thread saying it "*does* the exact same thing".

So I guess, you didn't clarify what you meant by "does" when you said "does the exact same thing". Thanks for clarifying that by "does" you mean "binds to the same ids".

Quoted 6 lines

and it's
author is getting annoying by refusing to answer my emails, but still,
please work with him to get the in-kernel driver to work up to your
needs.

Why are you pushing him to this when he's getting annoying to you? When he's not responding to your emails? Is he your friend? (perhaps more?--none of my business) Do you guys have beers at LinuxCon or other Linux conferences? Is he in the technology sector? Or is he just a "linux kernel engineer"? A guru in USB? Or maybe SCSI? Is his employer, Intel, pressuring you? Because when you removed all possibilities, however improbable, must be the truth. (after all they did write the XHCI HCD) All I see is a lot of nepotism, an inferior coding, inferior terminology used, inferior naming convention, incompetence in USB and SCSI protocols, etc, etc, etc.

Were you not part of the same thread where YOU RESPONDED addressing its author about his cowardly and lacking-professional-integrity ways of changing my patches including changing the commit messages without reporting what's been changed, how it's been changed and where it's been committed to? To refresh everyone's memory, here:
http://marc.info/?l=linux-usb&m=129133499714636&w=2
I guess you had to do that to be released of some kind of binding?

Do you not see HOW DIFFERENT the two drivers are? Do you not see the difference in quality, presentation, etc, etc.

Linux is truly open source, but closed community. You have shown in this thread that it's not about the code or the technology, but the turf. After I listed the technical differences between the two drivers in this thread (18 and counting), after you've seen the code and leading on with this thread despite the problems you're having with him and his not responding to your emails.  Why you are protecting his interests so blatantly despite the problems you're having with him is beyond me. (other than, again, if his employers is involved and pressuring you)

"...whatever is left, however improbable, must be the truth."

Quoted 3 lines

If you have problems with the maintainer, please let me know and I'll
see what I can do.

What? Were you not on the CC list of the patches I posted, did you not state in those threads you're having problems with him too? You yourself admitted above that you also have problems with him in those very threads. You are protecting a person, not code. Here are the threads:
http://marc.info/?t=128821112700001&r=1&w=2
http://marc.info/?t=128829652800022&r=1&w=2
http://marc.info/?t=128839891500022&r=1&w=2
http://marc.info/?t=128901886000001&r=1&w=2

Quoted 1 line

But again, this driver is not acceptable as-is, sorry.

"as-is"? You are giving me a clue, Greg! (Since you're NOT specifying any technical reasons.)

Greg, (to expose the obvious agenda and interests being protected) how about if I added his name as an author to my driver? Will you accept it then? Or perhaps I can send in a patch ripping out everything from the still-born uas.c other than the copyright and MODULE_AUTHOR() and substituting in the contents of my uasp.c (removing my name altogether)?

After all, people and the industry are learning a lot from this thread.  They are learning that the open source Linux kernel has been institutionalized between a few chosen people, regardless of capability, thus creating a closed community whose interests are highly protected. It's not about the code or the technology any more.

I think people need to know.

     Luben

#1 lore raw
Quoted 1 line

I think people need to know.

I think they will judge themselves from the full discussion. I'm not sure they'll vote in your favour however.

"So to expose the obvious, hypothetically, one of the resident "linux kernel engineers/maintainers" could've submitted AN EMPTY DRIVER that "binds to" the UAS id, and from then on, any other improvements would have to go through them"

Strangely that doesn't work because the community isn't full of idiots, in fact we actively select against them because anyone who is maintaining a driver needs to be able to deal with other people (at least other people who wish to be constructive).

"Do you not see HOW DIFFERENT the two drivers are? Do you not see the difference in quality, presentation, etc, etc."

I find the presentation *very* different. I'm rather worried about the manner in which it is being presented. Your driver may be the best on the planet - who knows - but if your response to needing to work with people is to accuse them of being part of some giant conspiracy or make offensive comments then that's rapidly going to outweight the quality of the code and you'll simply run out of people to talk to.

It's actually pretty hard to get in GregKH's kill file, and I suspect by now you are in a few others as well. So perhaps you can find someone working with you who has people skills and can sit between you and Greg and other maintainers and make progress ? It won't be the first time a user interface problem has been fixed that way in the Linux world.

But basically it's really simple. When we have an existing driver we work from it - step by step from where we are, to where we want to be. That series of steps should tell a story so anyone reading the patch series can understand how it unfolded.

We also have a process for dealing with people not responding, and irritating them enough they killfile you and refuse to deal with you isn't that process, especially when you then get yourself killfiled by the next maintainer up as well.

When you come along late and say "I've got this great other driver", then sorry you missed the party - you could have submitted yours earlier and the question becomes "how do I make the existing driver at least as cool as the one that was too late" not "how do I replace it"

Alan

#2 lore raw

Alan, don't you work for the same company as do the authors listed in "uas.c"?

Quoted 4 lines · Alan Cox

I think they will judge themselves from the full
discussion. I'm not sure
they'll vote in your favour however.

Alan, you're priming up here what people should think.

Quoted 10 lines

"So to expose the obvious, hypothetically, one of the resident "linux
kernel engineers/maintainers" could've submitted AN EMPTY DRIVER that
"binds to" the UAS id, and from then on, any other improvements would
have to go through them"

Strangely that doesn't work because the community isn't full of idiots,

I didn't say it is.

Quoted 5 lines

in fact we actively select against them because anyone who is maintaining
a driver needs to be able to deal with other people (at least other
people who wish to be constructive).

Did you not see the problems Greg has had with Matthew, your co-worker, and stating so? Let me refresh your memory:
http://marc.info/?l=linux-kernel&m=129133499814639&w=2

Quoted 7 lines

"Do you not see HOW DIFFERENT the two drivers are? Do you not see the
difference in quality, presentation, etc, etc."

I find the presentation *very* different. I'm rather worried about the
manner in which it is being presented.

Wait a minute... So a commit patch is not enough any more? Code is not enough anymore? Quick and knowledgeable responses are not enough anymore?

What was the "uas.c" presentation? ("uas.c" is written by a co-employee of yours.) Can you point me to it?

But you make a very good point: "presentation" is all there is to anything. If you've got the right presentation you can convince people of *anything*, literally anything.

For example, your response to which I'm replying is one such "presentation".

Quoted 3 lines

Your driver may be
the best on the
planet

Well, I don't know Alan? Is it? You can read code, right. No ONE has commented on the code, on technical matter at all. Instead Greg snipped it all out.

Quoted 5 lines

- who knows - but if your response to needing to work with people
is to accuse them of being part of some giant conspiracy or make offensive
comments

Wait a minute here... Is it about code and functionality then? Because if it is I didn't see any technical comments on the drivers.

Quoted 3 lines

then that's rapidly going to outweight the quality of the code
and you'll simply run out of people to talk to.

Here you're involving intangible things, but i realize it's part of your presentation.

Quoted 1 line

It's actually pretty hard to get in GregKH's kill file, and

It's just a move on his part to show he doesn't want deal with this anymore. Clearly he's expressed an opinion of having hard time working with uas.c's author, co-employee of yours. Clearly he's shown he doesn't want to comment on the technical matters, code, etc, despite all this he's shown to protect defunct code, written by a co-employee of yours, with comments in it like this:

Quoted 1 line

/* I hate forward declarations, but I actually have a loop */

and:

Quoted 5 lines

/*
* Why should I request the Status IU before sending the Command IU? Spec
* says to, but also says the device may receive them in any order. Seems
* daft to me.
*/

Anyway, I've noted these and other technical matter in the 18 major differences between the two drivers here:
http://marc.info/?l=linux-usb&m=129185378612218&w=2

Quoted 2 lines

I suspect by
now you are in a few others as well. So perhaps you can

You are telling people here what they should think and do.

Quoted 4 lines

find someone
working with you who has people skills and can sit between you and Greg
and other maintainers and make progress ?

There is no agenda for Linux and there has never been. My submission was purely to help others and give them something useful they can use to test and develop their UAS devices, an alternative to Windows. It took me two days to write this driver. I decided to help people because I could. There never has been an agenda to have a UAS driver (let alone mine) in Linux. But it was interesting to see the dynamics involved.

Quoted 4 lines

It won't be the
first time a
user interface problem has been fixed that way in the Linux world.

Or the last. Linux is open source but closed community.

Quoted 4 lines

But basically it's really simple. When we have an existing driver we work
from it - step by step from where we are, to where we want to be.

The existing driver shows lack of knowledge of UAS and SCSI. It falls short of almost everything. For the details see the 18 points here:
http://marc.info/?l=linux-usb&m=129185378612218&w=2

Why would you want to keep this in the kernel and "start from it", when we can rip it out, put the working and correct uasp.c in and then everyone including Matthew can contribute to uasp.c. It doesn't seem he works with the technology as I haven't seen any new patches to uas.c.

It's quick and constructive process. Out with the defunct and and in with what works. It's all about getting things done.

Quoted 4 lines

That
series of steps should tell a story so anyone reading the patch series can
understand how it unfolded.

But there would be NOTHING LEFT of the original driver. Just one patch like this I posted here to uas.c changed 86% of the code. A diff between uasp.c and uas.c shows 100% difference.

Quoted 2 lines

We also have a process for dealing with people not responding, and

"Not responding" -- are you talking about your co-worker?

Quoted 6 lines

irritating them enough they killfile you and refuse to deal with you
isn't that process, especially when you then get yourself killfiled by
the next maintainer up as well.

Quoted 7 lines

When you come along late and say "I've got this great other driver", then
sorry you missed the party - you could have submitted yours earlier and
the question becomes "how do I make the existing driver at least as cool
as the one that was too late" not "how do I replace it"

Alan, none of this is really important. As technologist, I thought it would be helpful and useful to write a functional UAS driver. Since the one in the kernel, was defunct, judging from how it behaved, how it locked up, how it went into an infinite loop, and the comments there-in. I did submit patches to uas.c, which got secretly modified, both the commit message and the patch, but never posted publicly as to what was modified. Even Greg commented on this not-so-good practice (see link above).

The uasp.c took two days to write.

But here is what I would've done had someone, especially a technologist, posted a working driver while mine didn't do the job: I would not have been quiet about it but simply asked to include the better working driver and I'd contributed whatever I could to it. It's a quick, practical and constructive thing to do. It's what gets things done. In the long run though "uas.c" would look exactly like "uasp.c" looks today. There is just so few ways to do the same thing.

However this isn't the first time this will happen or the last. rlt8139.c and 8139too.c come to mind, although the former had been around for a long time and there were plenty of devices out there for it. There maybe other instances like this.

Luben

#3 lore raw

On Fri, Dec 10, 2010 at 06:26:56PM -0800, Luben Tuikov wrote:

[Reflowed your text into 80 columns, you might want to look at your MUA configuration here.]

Quoted 6 lines
Quoted 4 lines · Alan Cox

"Do you not see HOW DIFFERENT the two drivers are? Do you not see the
difference in quality, presentation, etc, etc."

Quoted 7 lines
Quoted 3 lines

I find the presentation *very* different. I'm rather worried about the
manner in which it is being presented.

Wait a minute... So a commit patch is not enough any more? Code is not enough anymore? Quick and knowledgeable responses are not enough anymore?

The issue here is with the kernel change and risk management processes rather than the code.

Your new code adds a driver which replicates the functionality of an existing driver. We've had multiple implementations of the same functionality in the past. Usually what happens is that users and distros get confused and end up swapping randomly between the two different implemenations trading off the different bug and feature sets, which doesn't make anyone happy and there's a general idea that we should try to avoid that.

This means that the 1000 foot review comment is what Greg has been telling you - the standard approach is to work on the existing code in place, incrementally making it better. This avoids the problem with bug tradeoff (as there's only ever one version in the kernel at once) and makes it much easier to isolate any new problems if they are introduced.

Sometimes this isn't possible or a good idea for some reason, in which case the change should really explain that in the changelog (usually everyone involved will have some awareness of the issues already but a summary is useful for people picking up a new kernel release or similar). At the very least proposing such changes needs to involve some discussion of why a rewrite is required, and there needs to be some sort of plan for how everything should converge back onto a single implementation again.

Quoted 1 line

http://marc.info/?l=linux-usb&m=129185378612218&w=2

This is a good summary of what improvements the new driver brings (ideally more of it would have gone in the changelog), the missing bit is an explanation of why these issues can't be addressed with the usual process of incremental improvements to the existing code, discussion of how the existence of the two separate implementations would be resolved and discussion of the user visible impact of swapping to a new implementation.

Bear in mind that all your changelog said originally was:

| UASP: USB Attached SCSI (UAS) protocol driver

| This driver allows you to connect to UAS devices | and use them as SCSI devices

which doesn't say much more than that there's a new implementation of this.

#4 lore raw
Quoted 47 lines · Mark Brown

On Fri, Dec 10, 2010 at 06:26:56PM -0800, Luben Tuikov wrote:

[Reflowed your text into 80 columns, you might want to look at your MUA configuration here.]

Quoted 6 lines
Quoted 4 lines · Alan Cox

"Do you not see HOW DIFFERENT the two drivers are? Do you not see the
difference in quality, presentation, etc, etc."

Quoted 7 lines
Quoted 3 lines

I find the presentation *very* different. I'm rather worried about the
manner in which it is being presented.

Wait a minute... So a commit patch is not enough any more? Code is not enough anymore? Quick and knowledgeable responses are not enough anymore?

The issue here is with the kernel change and risk management processes rather than the code.

Your new code adds a driver which replicates the functionality of an existing driver. We've had multiple implementations of the same functionality in the past. Usually what happens is that users and distros get confused and end up swapping randomly between the two different implemenations trading off the different bug and feature sets, which doesn't make anyone happy and there's a general idea that we should try to avoid that.

This means that the 1000 foot review comment is what Greg has been telling you - the standard approach is to work on the existing code in place, incrementally making it better. This avoids the problem with bug tradeoff (as there's only ever one version in the kernel at once) and makes it much easier to isolate any new problems if they are introduced.

Sometimes this isn't possible or a good idea for some reason, in which case the change should really explain that in the changelog (usually everyone involved will have some awareness of the issues already but a summary is useful for people picking up a new kernel release or similar). At the very least proposing such changes needs to involve some discussion of why a rewrite is required, and there needs to be some sort of plan for how everything should converge back onto a single implementation again.

Quoted 1 line

http://marc.info/?l=linux-usb&m=129185378612218&w=2

Quoted 7 lines

This is a good summary of what improvements the new driver brings (ideally more of it would have gone in the changelog), the missing bit is an explanation of why these issues can't be addressed with the usual process of incremental improvements to the existing code, discussion of how the existence of the two separate implementations would be resolved and discussion of the user visible impact of swapping to a new implementation.

When done immediatly there won't have been a "user visible impact" as uas was only accepted a few weeks earlier than uasp, to be releases with 2.6.37.

As uas has not been in a release-kernel, by definition, there is no impact either way. No regressions or anything else.

If uas is reverted (or disabled like fanotify was for 2.6.36) NOW, there is time to resolve the matter for 2.6.38.
If uas is of the quality that Luben describes, it isn't fit for release with 2.6.37 anyway (or only with an EXPERIMENTAL tag, which it hasn't and doesn't depend on).

As a future user of UASP (The protocol) i'm happy to spend a few more weeks with UMS if it means the driver for UASP is really ready for transferring TBs of data when released.
(Altough i'm guessing that i will spend some more time with UMS anyway, as neither the packaging nor the website for my USB 3.0 enclosures mention UASP)

Bis denn

Signature · 4 lines
Real Programmers consider "what you see is what you get" to be just as 
bad a concept in Text Editors as it is in women. No, the Real Programmer
wants a "you asked for it, you got it" text editor -- complicated, 
cryptic, powerful, unforgiving, dangerous.
#5 lore raw
Quoted 10 lines · Matthias Schniedermeyer

As a future user of UASP (The protocol) i'm happy to spend a few more
weeks with UMS if it means the driver for UASP is really ready for
transferring TBs of data when released.
(Altough i'm guessing that i will spend some more time with UMS anyway,
as neither the packaging nor the website for my USB 3.0 enclosures
mention UASP)

Yes, it is ready. If you have a UAS device, give it a try. The uasp.c driver is in the free-linux branch off of master in my github repo. See the announcement here:

http://marc.info/?l=linux-scsi&m=129227488508299&w=2

Luben

#6 lore raw
In reply to Matthias Schniedermeyer

I was just chilling down late in the evening by reading a short LKML overview, and then inadvertently managed to hit this bomb shell ;)

Quoted 2 lines

As uas has not been in a release-kernel, by definition, there is no impact either way. No regressions or anything else.

Since I've actually just been oldconfig'ing uas driver yesterday on a new installation, I can relate to UAS kernel support being quite new in its entirety.

If this is the case, then I'd also tend to believe the situation is very different from other more painful (since longstanding) driver conflicts (e.g. 8139, some raid drivers, and also e100 as witnessed by myself).

The author of UASP might want to adjust some of his writings (especially the more personal parts),
however several parts of his criticism (e.g. no focus on code review) seem valid from my POV.

If the new driver indeed is a whole lot better than the other newly submitted/unreleased driver, then I'd fully welcome a reevaluation of the situation.

For the author of the original submission this might of course be a "less than positive" (to put it extreemely mildly) situation. Still, I'd hope that there could be sufficient agreement on how to proceed and how to make sure to have the most fitting code get into the kernel and be maintained properly, ideally by continued interest by _both_ authors (or perhaps via "personality firewalls" ;), obviously.

Alas, I'm feeling like I'm stating obvious blathering here.

Anyway, by now you know which direction I'm tending in, despite the not terribly warm writings by a certain author ;)

Disclaimer: no code review (comparison) performed, sorry. Disclaimer #2: as the author of a still-outstanding kernel patch (uhm - hi Greg ;) I'm in a sufficiently unfavourable position to comment on this...

Andreas Mohr

#7 lore raw
Quoted 10 lines · Andreas Mohr

I was just chilling down late in the evening by reading a short LKML overview, and then inadvertently managed to hit this bomb shell ;)

Quoted 2 lines

As uas has not been in a release-kernel, by definition, there is no impact either way. No regressions or anything else.

Since I've actually just been oldconfig'ing uas driver yesterday on a new installation, I can relate to UAS kernel support being quite new in its entirety.

Quoted 8 lines

If this is the case, then I'd also tend to believe the situation is very different from other more painful (since longstanding) driver conflicts (e.g. 8139, some raid drivers, and also e100 as witnessed by myself).

The author of UASP might want to adjust some of his writings (especially the more personal parts),
however several parts of his criticism (e.g. no focus on code review) seem valid from my POV.

I think you may be missing some code review on the linux-usb mailing list. Luben posted his original patch to the uas.c driver there, but he included behavior changes, code style modification, and variable renaming in one giant patch. It was very difficult to wade through, so I asked him to break the changes into a series of patches (perhaps 18, since he has 18 points?).

I did attempt to give a partial code review of Luben's original patches, and he has yet to respond to any of my points:

http://marc.info/?l=linux-kernel&m=129021816023869&w=2

I also asked him to change a different patch for his driver, to move a work-around into a better place:

http://marc.info/?l=linux-kernel&m=129192831709856&w=2

I have yet to see an updated patch set, only personal attacks.

Sarah Sharp

#8 lore raw
Quoted 3 lines · Sarah Sharp

I think you may be missing some code review on the linux-usb mailing
list.

There was never a review of uasp.c. Let's be clear on that. Because the way you've written the above sentence, someone may misread it to understand that your co-workers and Greg had written extensive review of uasp and had rejected it based on some tangible virtue. Greg's NAK came 4 minutes after I posted it and I mentioned that in my response (in this thread).

Quoted 5 lines

Luben posted his original patch to the uas.c driver there, but he
included behavior changes, code style modification, and variable
renaming in one giant patch.

I wanted to rip the whole thing out. That patch changed a modest 86% of the code and I did mention that in the patch:
http://marc.info/?l=linux-usb&m=128901883420388&w=2

Quoted 2 lines

It was very difficult to
wade through,

That's why it's better to rip the whole thing out and put in something properly done.

Quoted 4 lines

so
I asked him to break the changes into a series of patches (perhaps 18,
since he has 18 points?).

I'm not going to generate 18+ patches that changes 100% of that driver into a completely different and working driver.

First, I've no time for this. Second, it's much better, quicker and error-proof to present a whole _new_ piece (and show how it's done), moreover the fact that there are no UAS devices out there at the moment.

You and anyone else can certainly contribute to uasp.c. What's the point in converting the defunct uas.c to look like the working uasp.c (two different drivers)? The proper thing is to pull uas.c out and put uasp.c in (working and properly implemented (see my original editorial and patches to uas.c)).

Quoted 5 lines

I did attempt to give a partial code review of Luben's original patches,
and he has yet to respond to any of my points:

http://marc.info/?l=linux-kernel&m=129021816023869&w=2

Now that you asked, I just did respond to this. The reason I didn't respond to it before is because I didn't think there was any point in responding then, as I do now. Back then I would've said: the whole thing should be ripped out and re-written. Now I'd say: refer to uasp.c on how it's done. It's available on github.

Quoted 5 lines

I also asked him to change a different patch for his driver, to move a
work-around into a better place:

http://marc.info/?l=linux-kernel&m=129192831709856&w=2

And I responded to this in the next entry in the thread in the same link. The parameter still has value. Most likely it wouldn't have to be set (but it still has value in the very very rare case...).

Quoted 2 lines

I have yet to see an updated patch set, only personal attacks.

A patch set for what? UAS? See uasp.c on how it's done. For the host that mis-calculates the number of streams it reports in HCCPARAMS? I still haven't had time to plug in a newer version of this HC. When I do and if it still misbehaves, I'll send you a dump of the PCI config space and you can decide what you want to do.

Luben

#9 lore raw
In reply to Mark Brown
Quoted 14 lines · Mark Brown
Quoted 1 line

--- On Fri, 12/10/10, Alan Cox <[email protected]>

Quoted 1 line · wrote:
Quoted 1 line

"Do you not see HOW DIFFERENT the two drivers

are? Do you

Quoted 4 lines
Quoted 4 lines

not see the
difference in quality, presentation, etc, etc."

I find the presentation *very* different. I'm

rather

Quoted 4 lines
Quoted 2 lines

worried about the
manner in which it is being presented.

Wait a minute... So a commit patch is not enough any

Quoted 8 lines

more? Code is not

Quoted 1 line

enough anymore? Quick and knowledgeable responses are

not enough

Quoted 1 line

anymore?

The issue here is with the kernel change and risk
management processes
rather than the code.

There are no UAS devices in the market. Risk 0. The in-kernel UAS driver is defunct, both in the way it is written, the terminology it uses and in the fact that it simply does not work. See the technological treatment here:
http://marc.info/?l=linux-usb&m=129185378612218&w=2

Quoted 3 lines

Your new code adds a driver which replicates the
functionality of an
existing driver.

"eplicates the functionality of an existing driver": Now read this again:
http://marc.info/?l=linux-usb&m=129185378612218&w=2

Quoted 3 lines

We've had multiple implementations
of the same
functionality in the past.

I've not followed the kernel lately, but I do remember what you speak of: rtl8139.c and 8139too.c. And this duplicate functionality was in fact a re-write (as the notice in the latter driver explains) WHILE there were hundreds of thousands devices out there. Why re-write instead of "contribute to the one in the kernel, patch by patch" as Greg and Alan would like you to believe? And at the same time WHILE there are (were) so many RealTek chips out there...

Quoted 9 lines

  Usually what happens is
that users and
distros get confused and end up swapping randomly between the two
different implemenations trading off the different bug and feature sets,
which doesn't make anyone happy and there's a general idea that we
should try to avoid that.

Ah, yes, I see: you're trying to _protect_ distros and end users. But the fact is that users and distros are smart enough to figure out what works and what doesn't work. People working at distros are smart enough to test and see what works. And often enough, distros would contain, new and/or rewritten drivers from the one in the mainline, especially for new or rare devices which their customers want support. So, please don't try to pull this bs of "we're protecting the end user and distros".

Quoted 5 lines

This means that the 1000 foot review comment is what Greg has been
telling you - the standard approach is to work on the existing code in
place, incrementally making it better.

Just one patch I submitted changed 86% of the driver:
http://marc.info/?l=linux-kernel&m=128901883520391&w=2

Adding them all up, it'll be 100%. uas.c should be ripped out altogether. Do it now, while there are still no devices out there.

Quoted 6 lines

This avoids
the problem with bug
tradeoff (as there's only ever one version in the kernel at once) and
makes it much easier to isolate any new problems if they are introduced.

The only problems I've seen so far are with uas.c.

Quoted 9 lines

Sometimes this isn't possible or a good idea for some reason, in which
case the change should really explain that in the changelog (usually
everyone involved will have some awareness of the issues already but a
summary is useful for people picking up a new kernel release or
similar).

See this: http://marc.info/?l=linux-usb&m=129185378612218&w=2

Quoted 7 lines

At the very least proposing such changes
needs to involve
some discussion of why a rewrite is required, and there needs to be some
sort of plan for how everything should converge back onto a single
implementation again.

uas.c doesn't work and is very badly written piece of code. uasp.c works and complies with USB, UAS, SCSI and Linux kernel code programming guidlines. Again see this:
http://marc.info/?l=linux-usb&m=129185378612218&w=2

Quoted 1 line
Quoted 1 line

http://marc.info/?l=linux-usb&m=129185378612218&w=2

Ah, I see, so you have it.

Quoted 5 lines

This is a good summary of what improvements the new driver brings
(ideally more of it would have gone in the changelog), the missing bit
is an explanation of why these issues can't be addressed

Quoted 8 lines

with the usual
process of incremental improvements to the existing code, discussion
of how the existence of the two separate implementations would be
resolved and discussion of the user visible impact of swapping to a new
implementation.

I'm not sure what needs to be resolved: rip out the defunct uas.c and put in the working uasp.c. Then move on with more interesting things (like one's real job).

Quoted 10 lines

Bear in mind that all your changelog said originally was:

| UASP: USB Attached SCSI (UAS) protocol driver

| This driver allows you to connect to UAS devices | and use them as SCSI devices

which doesn't say much more than that there's a new implementation of
this.

It doesn't need to say any more, if the defunct uas.c is ripped out. This isn't a "why uasp.c should be in and uas.c out"--it's "here is a working UAS driver".

Luben