Bus Supplemental Application: Fill & Download for Free

I will list some of my favorite Mandarin Chinese blogs, but I also want to stress that these blogs and other resources will help you supplement your Chinese learning, not teach you Chinese.Learning Chinese takes a lot more than self-study and an interest in China. You really need personalized feedback throughout each step of your journey, especially the first steps.Here are some of my favorite online resources:Chinese Breeze: This site is well-designed and helps you learn some Chinese tips and hacks. It is applicable to those learning Chinese from around the world in various locations.Chinesepod: This site is a treasure trove of resources, from audio to video.ChineseGrammarWiki: This site has thousands of useful Chinese grammar resources.Chinese Hacks: This site as lots of Chinese resources related to gaming and popular online culture. You never quite know what you’ll find next.East Asia Student: Lots of links to translations and cultural texts which can supplement your learning.In addition to these aforementioned blogs, there are some apps that can help your self-study as well (on the bus, before bed, waiting on dessert, etc.):Anki, a flashcard system. Anki is an “intelligent” program, meaning the more you’ve mastered a flashcard, the less it’ll show up in your deck. This is a good, progressive way to study.Duolingo and Memrise are becoming more popular options for learning Chinese and other languages when you’ve got free time.Pleco is a dictionary that will help introduce and improve with Chinese writing, in particular.Skritter is also great for memorizing and practicing writing characters. Although speaking/listening is more important in the beginning, you may find the characters offer an interesting glimpse into the Chinese culture and way of thinking.You should consider these as supplements to your Chinese progress, not a foundation.To truly learn Mandarin Chinese, you need feedback and guidance from a qualified Chinese teacher. For me, it was online Chinese classes that laid the foundation for my learning progress with Chinese.Along with supplements, learning Chinese online and guided self-study is an excellent base for Chinese language learning. Stay motivated!

Thread consortium/alliance is really focused on a mesh networking protocol for the 802.15.4 Radio standards. AllSeen and OIC are also solving pieces of the puzzle. MQTT, XMPP, CoAP/6LoWPAN, DDS etc. are the many standards useful to various IoT applications. AllSeen is really leveraging the D-Bus protocol/mechanism to implement a proximal network of things. OIC (assuming it is based on Intel® Common Connectivity Framework) leverages XMPP.(BTW we don't know much about Thread today - but given that its focus is "IP" over 802.15.4 - it would be interesting to see how it co-exists or complements or supplements 6LoWPAN/CoAP)

While all the answer’s here are good, I think I have a relatively unique background to answer this, having worked “both sides of the fence” on projects with similar goals. I have written many programs related to regular expressions, parsing, and those forms of pattern matching. I also was the lead architect of a chip that Intel manufactured that solved the problem in hardware. So, I have seen the same problem solved both ways.So, let’s start with the first part of the question. These days designing a chip is generally done in a language like Verilog or VHDL. While there are some nuances in those languages, for most purposes they are programming languages. In fact, my first project at Intel was writing a Verilog compiler that output C code to do very fast simulations of chips as they were being designed.That doesn’t mean you treat them (Verilog and VHDL) like programming languages. You don’t just take an algorithm that you have some C code for and compile it and magically get a chip out. There are some vendors that promise that, but it is a vacuous and misleading promise. What makes a good program is very different from what makes a good hardware design. Very different. I’m not sure I can even explain what that difference is. It’s like comparing soccer and gridiron (what the world calls football and what the US does) and saying that they both descended from the same sport so they must be alike. Errr, no.And that’s despite the fact for a problem like regular expressions one builds an abstract machine and writes an interpreter for it and a compiler that takes what patterns the user wants to match and creates a program that the abstract machine’s interpreter runs. Thus, both a hardware and a software solution to the same problem use essentially the same solution technology, but the results at the implementation level, and the things which make a good solution are different.Again, I cannot really explain that to someone who hasn’t done both. But in hardware, you get parallelism (and very fine grained parallelism, each gate is essentially parallel with every other gate) for free—in fact, you have to work very hard to limit it and make it sequential. In software, sequentialism is free, you have to work hard to make it parallel. Threads and locks and semaphores are all non-trivial and it is almost never fine grained parallelism, vector instructions (SIMD code) being a very small exception.So, in that way designing hardware and designing software is not alike. Not in how you make a good solution. Even if you are solving the same problem, roughly the same way. The devil is in the details as they say, and the details are very different.But, one of the other writers mentioned the productization process. Making a chip and making software are also very different at that level. Making a CPU chip or something of similar complexity takes about 5 years from idea to actual fabrication and another couple of years before that chip is actually deployed in computers or other devices. It also takes a significant number of people with wildly diverse specialties. I suspect in the end several hundred people worked on “my” regular expression chip and that’s just at Intel. If one of our major customers had built it into one of their products, you might have added close to another hundred. The software startup I did which did a similar product in software, took two people for most of its life. We spent nine months from the version sketched out in one night to our first commercial version (and that’s because we changed our plans mid-stream, we could have had a commercial version in three months if we had planned on doing it from the beginning and not a different larger project).Let’s next talk about the process where one fixes errors. In software, if there is a problem, you can often find a solution in a few days and have a fix for it out in a week. Moreover, you can ship the fixed solution to every existing customer if you want, especially these days where you are shipping code over the internet anyway. Plus, there is no “inventory” of outdated versions you have to scrap. The cost of fixing software is not physical.Hardware is a very different story. There was a small but crucial bug in a part of the regular expression chip in the A0 stepping, not in our code but in code we depended on that hadn’t been written carefully and did not account for the fact, that we were both reading and writing on a particular bus. All the other parts of the chip only did read operations on that bus. It was broken, but only for us. The next stepping B0 did not occur for another 9 months and all the A0 chips were useless for our application. This was a two line fix to that particular part, but it took 9 months of waiting to get it.Moreover, if you look at the production costs for hardware versus software, you see the same difference. Building our software cost us a pair of computers. The original one was a $1k Atari ST, but we replaced that with a $15k Sun Workstation and a $2k PC, because we could afford to (and there was no market for an Atari version written in Pascal). Manufacturing a chip requires a FAB that costs upward of a Billion dollars. That’s 5 orders of magnitude difference, a 100 thousand times more expensive.How do these differences in cost affect the development process? With software, the main cost is “opportunity cost”. If you don’t get your product out to the market in the right window, someone else will take your market. Our product faced that with ANTLR a free product developed at roughly the same time. In the end, the free product took over the market even though there were reasons why for certain use cases, ours was better. So, developing a product quickly often is a priority. Moreover, once someone sees your idea, numerous variations will spring up. Although, it isn’t just software, look at Uber v. Lyft et al, nothing ties people to Uber other than inertia. Software is like that. Ideas are easy. Turning them into products and getting people hooked on them versus your competition is the hard part.The result of that is the “time-to-market” is often everything in a software development process. If you can cut a corner and skip a step, but get a minimally functioning product out the door, you can “win the lottery”. So, investment in quality software products tends to be skimped upon. Moreover, because you can easily fix it, the ramifications of any short-sightedness of such choices can be repaired, often at little cost. As long as the fix isn’t a public relations disaster. The technical problems facing Zoom to make it more secure are not the issue. Their issue is loss of trust by the marketplace.Now, look at a similar problem for the hardware point of view. The two line fix to a usually unimportant part, cause a 9 month delay. Worse, no customers were able to try the product during that delay. So, all the design wins that should have happened by getting them to adopt it were lost. That meant that all those customers had to choose an alternate solution. Thus, much of the 5 year effort to develop that chip were lost. Not because of a problem in the part that was built, but due to not testing another part properly and assuming it was going to work. Again, if that had been software, maybe it would have taken us a week to fix.So, what do the hardware people do that software people don’t?First, of all, they test and test and test. There were at least three levels of validation between our chip design and putting it into the A0 stepping. All those unit tests that people complain about writing as useless and trivial. Well, the hardware people do them (except in the case of the bus design they didn’t). Moreover, they supplement those unit test with random tests and assertions. Does you unit test work? Good, what if we vary the conditions it is tests under? What if we try inputs that your unit test doesn’t try? Not just a few, but billions upon billions of them. And that level of validation is just at the design level.Next they do integration testing. Does your part of the chip work with the parts that it is supposed to? Do your unit tests work with their unit tests? Do they work under the billions of input variations?And, after that, they do system testing. Does the chip boot? Can you bring up Linux (or Windows) on it? Once they get the OS to boot, do the intended applications run? Does your demonstration software run? What happens when it is running for hours and under various workloads? (That’s where they found the flaw in the broken part. If under enough stress, it overflowed a FIFO and dropped data going across the bus, because they didn’t write the check for the full condition into the write logic. A simple unit test would have caught that, but no one wrote it.)But with hardware, there are no easy fixes for those problems. Well, sometimes they can be worked-around with software, but not always. So, with hardware getting it right, getting it as close to perfect as you can, that’s important. With software, you just fix it.Well, not if it is avionics software, or software in a medical appliance, or software with legal conformance requirements. Then, you do take more time.But that isn’t the software most people are thinking about. That isn’t the software in your phone or PC that you get a patch for every month, because you cannot patch hardware. If it is broken, you have to physically take it apart and replace it (and if it is soldered in, that may be impossible or at least very difficult). When my phone died, they simply gave me a new one. Fixing the old one was not economical. If it was software, they just would have downloaded a patch.And, yes, that’s why I write about the importance of unit tests and assertions in software. I just wrote a trivial piece of code a couple of days ago. Turns a number into a string. 10 lines of code maybe. Worked fine out of the box. Well, except I forgot the case where the number was 0 and didn’t write a test for that case. So, I used it where I needed and sure enough, the next level up broke because the 0 case wasn’t handled. It didn’t have a unit test for that case either. But, you know what, now they both do. You write tests not because you are worried about what you know. You write tests because you are worried about what you might have forgotten. But at least this was software, fixing it was easy. And it made me think about some other things I might have not remembered.Code isn’t magically bug free. It works only for those things that you have taken care of, whether intentionally or not. Write your code intentionally and then check for what intentions you have forgotten. Then, you won’t be sending out so many fixes month after month. You still will have some, just fewer.The hardware people get that right, because it is too expensive for them not to. And, yet there are still flaws like Meltdown vulnerability that show that they haven’t thought of everything. And, no, you cannot just send out new chips every month to fix the problem.