USB Stack
Rick Murray (539) 13840 posts |
When the phone rings, I go into the other room and hide until the noises stop. 1 Not that I’m actually any good at that… |
Clive Semmens (2335) 3276 posts |
Hear, hear to all that! |
John WILLIAMS (8368) 493 posts |
I was going to take this to Aldershot, but on reflection feel that it’s far too important to hide away there: Rick said:
That is very well-put, and I shall be using it myself! Clive said:
which was really what I wanted to say! So Andrew’s view of the world he’s dealing with is very Andrew-centric and exclusive (in the perjorative sense): I have myself little left to offer the waiting world, but this is not true of Rick and Clive who have contributions of merit yet to make. I very rarely answer the ‘phone at all – usually only to my wife. As for the “mobile” ’phone, that stays firmly in the house unless I need a time reference or to take a photo, and happily it seldom rings, tho’ more recently as the NHS have the number now. Or, of course, write a message – a much more civilised and effective way of doing things! |
mikko (3145) 123 posts |
From here it looks like Andrew does a fine balancing act of supporting RISC OS and supporting his family. Whether or not the the OP gets in contact with him, I hope the work goes ahead and that we all see the benefits of an improved USB stack. |
John Jeffords (8738) 26 posts |
Yes, there is. And some of it has been resolved by the things that have been said on this forum, and some have not. For example, I still can’t see why RComp get preferential access to Iris and are allowed to rebadge it for their own systems. HOW is this fair on other vendors? Unless other vendors are given the same access, this gives RComp a clear market advantage. It consistently feels like Iris is an RComp product, not a RiscOSDev product. Back to the topic, then, and given that RiscOSDev like to “pre-announce” what they’re doing, what is RiscOSDev’s plan for the USB stack and how does it differ from what the Bounty asks for? If RiscOSDev and ROOL are working collaboratively, then surely the USB bounty should reflect both sets of needs. Otherwise it’s duplication of effort, having two people attempting different things with the USB stack. As @ Rick Murray accurately said:
I am lead to believe by someone who attended the ROUGOL Awards show that this was a nomination for the Broken Cog Award too? So it clearly isn’t just me who is concerned by this. For me, these points also echo my concerns. @Richard Walker said
@Peter Howkins said
@Rick Murray said
The whole Iris thing, upon reading back and watching videos, feels exactly like this. “Shareholders” or whatever they are, had to sign NDAs, I understand: “We’re going to build a browser, but don’t tell ayone”? I am worried that any USB stack work outside of the bounty could end up going the same way. Hence my original posting on this topic. @Andrew Rawnsley said
That’s great, but it means nobody else knows a jot of what’s happening. And that may deliberate, or it may be unintentional, but it does mean the essence of that conversation is lost to others who are interested. I’m sure you don’t want to have the same coversation multiple times a day to help clarify things to lots of interested parties. Perhaps get your web designers to use the News section of the riscosdev.com website properly. Do people really want to know that there were some shows in 2019 and 2020?
Which is my point over Iris. An R-Comp release. Why do R-Comp get to release a separate version? Do others? As I am repeatedly told, RComp and RiscOSDev are separate companies, so why does this happen? On a conciliatory note: @Andrew Rawnsley said
I’m not sure what you mean here. If you wish to clarify, that would be helpful.
And I can only say that I am sorry that your mental health is negatively affected. That is not good for you, your family or anyone.
I was reluctant to make that point initially, because I didn’t want to create a self-fulfilling prophecy. But the “poaching” approach created some concerns for me that it felt like sidelining. I don’t want to damage RiscOS, as I know you don’t. @Richard Walker said
That is very much appreciated. I think it is fair to say he has, with a number of others, significantly helped keep RISC OS alive for the past decade. Thank you. BUT we’re all still not looking at the bigger picture, which is keeping the platform alive. This needs new users. We had or have the biggest opportunity to grab a whole load of new users, with the Raspberry Pi. Currently, they can buy a very capable RiscOS machine for £35, add an SD card with the ROOL image and they’re good to go. And they want to get on the Internet, so load up Netsurf and, as @Rick Murray put it, they’re back to 1998. They then see that a browser is available that does much more modern stuff, but unless they buy some different hardware (from one particular vendor, but that’s not my point here), they have to pay £49 to get access to it! More than they paid for the Raspberry Pi for something that people, on almost all other platforms, expect to be free. After thirty years of using RiscOS, I accept it’s shortcomings, but new and potential users won’t, and will move on, and we need them to stay. Just release Iris to all and be done with, give the source to ROOL to publish (or publish it yourself), and let other skilled and experienced users contribute. |
Paolo Fabio Zaino (28) 1882 posts |
IMHO, this is a huge problem for the community, even for people that are just contributing by coding 3rd party apps and tools. No one knows what the next change delivered will be (unless one follows the nightly builds, meetings of various clubs, bounty state, search for any update from developers that may happen in the forum here or on twitter or on a 90s looking website called xyz.unibroken.co.mars/~/wth). I though ROOL website was the official place where to follow progresses and ideas being discussed/developed by both ROOL and ROD. So, what could be done to help this? What follows are just some ideas (maybe they could be helpful): 1) What about a page with updates of a meeting on this website? “In today’s meeting we have discussed X, Y and Z, what ya think about it? Comments below (no jellyfish allowed please).” 2) What about a podcast discussing or explaining things? Podcasts are audio only and could be (relatively) easily used on RISC OS (for the “I have RISC OS only as it is the only path forward!” kind of users) – For the discussions that may contain NDA subject content you can edit the podcast before release it. 3) Youtube? 4) For the future (when the Open Source concepts have been absorbed and digested by all involved parties) what about an Open Zoom?
And, if we would like to match both (the ones that like the audio chat and the ones that don’t), RISC OS not being full youtube ready and trying to make it as low effort as possible, probably idea N.2 is the best compromise? Communication is key in open source projects. As always just my 0.5c. A clarification about Open Source Source Wikipedia: This numbers are growing year on year and there is a reason for that: Linux focuses on solving INDUSTRY problems or helping companies to solve them. This is actually how successful open source works. This is also why Linux is struggling to have a great desktop experience, because Linux Desktop doesn’t solves industry problems and so will not increase donations and support coming from the industry. 2nd, when it’s not about money then it’s about coding and again the industry compose a huge portion of Linux improvements: Source Phoronix: Check that link please, do not ignore it. Ok, for the ones that insisted in ignoring the link above: Corporations are increasingly investing on Linux not only with money, but with code and 2020 has seen numbers that would make Microsoft get pale (oh wait Microsoft is also a board member of the Linux foundation, platinum member!). This is why Linux struggles so much to offer a great Desktop experience, there is no interest in the desktop market. Then there is the Open Source totally-free or that use very little donations, the one that struggles to move forward in most cases. Why is it so? Because developers (believe it or not) do have to pay for food, rents, assurances, mortgages like normal people do. Yup, I am sorry to disappoint you, but we are normal people like everyone else. So, when Open Source is unfunded and unpaid, then it’s generally slow and rely purely on an hobbyist-level amount time/effort and pure dedication of people. But wait, isn’t one of the biggest complain in the RISC OS scene that progresses are too slow???? The obvious conclusion then: So, if RISC OS has to become the "Paladin of Desktop computing” then should we expect the “Desktop users” to become platinum members of ROD? ;) Disclaimer: I am not involved or affiliated with any RISC OS companies, ROOL and ROD. |
Andrew McCarthy (3688) 605 posts |
Good to have you back, Dan :) Dan asked for thoughts around USB. We seem to have drifted off-topic. It might help if the next off-topic poster created a new thread to allow the community to further air their concerns for RISC OS and the possible tensions around stakeholders invested interests affecting the future of the OS; Leaving this thread to the USB topic. EDIT: Thank you, John J. |
John Jeffords (8738) 26 posts |
I agree. I’ll copy my post off to another thread in, appropriately, “Community Support” called Open Working if Paulo wants to copy his follow up over as well? |
Doug Webb (190) 1180 posts |
Well hopefully he hasn’t been put of by another pointless them and us debate side tracking things, have we not learnt any lessons over the past 20 + years. Lets leave it in the past as everyone is trying their best in my view. I hope Dan is going to have a go at the USB stack as it needs it and lets leave him to do it how he sees fit and to release the updates when he is comfortable and if he wants others input then let him decide that. It is a big enough project on its own without others adding to the pressure. |
Chris Hughes (2123) 336 posts |
Who exactly is going to do this work, both ROOL and the ROD stuff is not done by paid staff, they do it in between earning a living to put food on the table, etc.. This is to me unnecessary extra work for little benefit. |
Rick Murray (539) 13840 posts |
I think a real question that should be asked is – how are applications going to talk to any newer USB stack? As I understand it, there are two methods on RISC OS 5 (DeviceFS file handles, and calling the hardware directly in some manner 1), and Simtec did something else. It would make sense to keep the current methods (as existant software uses them), but if somebody is interested in developing the stack, this might be the time to ask if there is anything that could be improved? If not, great, but… anything? 1 I use the DeviceFS method myself because it is simple and even works from BASIC. I’ve not had the time to examine the updated USB MIDI code to see how the other way works. |
Richard Walker (2090) 431 posts |
I am not sure there is any other way, Rick. I t is DeviceFS or bust, isn’t it? In USBJoystick I used OSFile with DeviceFS, then I discovered a small variation where you use BufferManager so you can read without OSFile, I think. But it is still DeviceFS. Maybe a sensible API would be one similar to Linux/BSD, to aid with porting software? I have no idea how close we are to that, or even if it makes sense. |
Rick Murray (539) 13840 posts |
Ah, yes, BufferManager was the other way, I think. So that’s the file way a little below files, then? I think the benefit of the current way is that it can be used from anywhere, including BASIC, that is often a second class citizen as it can’t do stuff like callbacks and what have you. I won’t comment in the Unix API, I’m not familiar with it. |
Kees Grinwis (3528) 18 posts |
Shouldn’t DHCP be done automatically without any involvement from the UI? When the network (either wired or wireless) is going from a disconnected to a connected state the TCP/IP layer should request an address via DHCP (when this is enabled of course). The manual action should only be an option in case this address request did fail. |
Clive Semmens (2335) 3276 posts |
If you’ve got any Apple devices on your network (such as someone’s iphone wifiing in) then failures (address clashes) are only too horribly likely. Reserved addresses for anything you know about are your friend… |
Steve Pampling (1551) 8170 posts |
Yes.
Apple devices aren’t too concerned about whether you reserved an IP for something else in the DHCP scope either. Apple devices seem to operate in ignorant mode. Reserve an IP for the Apple device and it might not clash. |
David Feugey (2125) 2709 posts |
I’m pretty sure that if I talk to ROD, I could have my own derivative of Iris to sell with my hardware too.
It would be so nice if it’s was so easy. RISC OS itself is open source. See the numer of ‘other skilled and experienced users’ who contribute. It’s perhaps incredible, but most of new developments are paid by someone (bounties or private contracts). The problem is complex. Pay for something and deliver it only when finished: people complain you don’t communicate on your work. Pay for something and communicate on the development: people complain source code should be available from day 0 and that you should interact more with users. I agree, but if we do all of this, we’ll simply not have enough time and money for the development work itself.
Correct. And time too. |
Steve Pampling (1551) 8170 posts |
What do you make of the statement that the number of commits is at its lowest since 2013 and the number of authors is falling too? People losing interest? |
Paolo Fabio Zaino (28) 1882 posts |
Time is a concern only when there are (well) time commitments. Time commitments usually engage when a job is paid for, not when it’s for free (as in free-beers). I am not against Bounties or even paid jobs, if that’s what people want. However, in a paid job documentation and tests are usually also a requirement. Now, the Linux Kernel (as an example, but so does BSD and many others) comes with full documentation details, RISC OS does not. So, to be absolutely clear here, the issue with the communication in RISC OS Dev Model is before, during AND after code is done, hence it’s a cultural problem not a time problem. Hope I am making sense. |
Paolo Fabio Zaino (28) 1882 posts |
One of the most active Linux Kernel part is the Drivers part, and that is mostly controlled by Companies (which is partially why you see the trend of less patches submitted and more from companies).
This one is a complex answer. First of all since the “colonisation” of the Linux Foundation by companies with big interests a lot of the Linux Kernel efforts have been focused on delivering kernel features that are addressed to solve companies problems, not average Desktop user problems. This combined with the majority of work done on the Drivers side and such a little interests in the Desktop use of linux has kind of created 3 different OSS worlds: The Linux Kernel folks, the Compiler folks and the Desktop (UI and App) folks. So we should probably analyse the shifts in all these 3 aspects. But Yes there is an “underground” dissatisfaction with the way the Linux Foundation is driving Linux developments. How much this counts it’s hard to say, because honestly also BSD has seen a lost of developers and slowed down a lot (all BSD distros). A possible extra explanation on this is the enormous interest created around the AI/ML/DL bubble (and yes it’s a bubble right now, not all applications of it are being successful), so many new talents are looking at AI/ML/DL instead of kernel developments. That also can be another source of reduced patches.
Of course there is also this, also because getting into the Kernel work is not for everyone and requires very good C knowledge, which is a language that is now seen as scary and that gives you only “maintainers” jobs also in companies. So yes I think a mix of these reasons are also an element. But surely there is a component of sustainability. Free as in free-beers does not exists, in the sense that if I spend time porting a library to RISC OS for no money, that time “sustainability” has surely been paid by some other activities. It doesn’t matter how Open Source fanatic one can be. If you have no pennies in your pocket whatsoever I am sure you’ll focus on something else that, at least, allows you to survive if not make you have a good life. And finally, the Linux Kernel architecture is so well made that do help to reduce the amount of work needed to fix things and/or try new approaches, so yes it is a fairly mature project indeed. |
Paolo Fabio Zaino (28) 1882 posts |
@ Chris Huges
Sorry Chris, just saw your comment. So, a post meeting “minute” (bullet points list of discussed things) it’s something they should be doing already, if they aren’t I find it hard to believe they can keep track of what they discussed and are going to do. So, I am not sure what extra work you are talking about, unless these meetings are completely informal, casual and amicable ala “Hey Andy, what’s up dude? Fancy some tcp/ip patch to bluethootey today? Btw do you know the joke about a USB Key, a Wifi Antenna and an SDCard go to a bar…” :) (I can’t imagine Steve Revil like that loool, but everything is possible in this world, who knows!) |
Rick Murray (539) 13840 posts |
Ah, but the thing is, we’re looking at it from the outside in. They may have clear communication, just choosing not to blab it all over the forums… I’ll give you a real world example. The company I work for is opening up new markets, devising new products, and we’re going to be working our asses off until Christmas. The first we know about any new product is when information on “make this” turns up on the factory floor, along with various people from R&D to oversee the process and check that everything is turning out right. Anyway, the people at the top clearly know what’s going on. From that point down, it falls under “a need to know basis”, and the guy wielding the bog brush does not need to know.
Not just that, but you need good knowledge of how a kernel does it’s thing, so as to be able to modify and extend without breaking stuff or more importantly compromising security.
Less “bubble”, more “bullshit”. You can’t have “intelligence” without understanding, and you aren’t going to get either with what amounts to sophisticated pattern matching. Why do you think facial recognition has so much trouble with black people? Everybody freaked out when Facebook identified some black people and offered to show other pictures “with primates” 1. It wasn’t intentionally racist, that bunch of pixels looked like some sort of monkey to the machine. Hell, I doubt it’s even capable of determining the difference between black skin and black fur because both are low contrast and both are likely to have been presented in training photos of varying degrees of crappiness. For sure, sometimes an AI can output something impressive, but your average three year old could crawl rings around it, because it’s just a machine following some complex equations. Follow-ups should probably be sent to Aldershot… 1 Actually, that’s correct. Humans (of any colour) are primates. Though such things are likely beyond the grasp of the average Facebook user. ;) |
Paolo Fabio Zaino (28) 1882 posts |
@ Rick
That’s not Open Source philosophy. So, it’s perfectly valid if a project is not open source. There have been few cases on Linux like that and even Linus Torvald complained about them. So yes you are confirming is a cultural issue, sorry. ok time to move this chat somewhere else, it has nothing to do with USB Stack :) |
mikko (3145) 123 posts |
This thread started out with the OP stating “I’m looking at refreshing the USB stack.”, which, coupled with the fact there’s now nearly £15k on offer for doing just that, is encouraging enough to resurrect this thread for. Are they still mulling this over, I wonder? |
André Timmermans (100) 655 posts |
Interesting: 29-Oct-2024 Handover to new developer, taken off hold |