Announcement

Collapse
No announcement yet.

Mozilla's Incredible Speech-To-Text Engine Is At Risk Following Layoffs

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • SystemCrasher
    replied
    The technologies that Mozilla was pushing forward with FirefoxOS were offline-first applications that used local browser storage APIs. The whole point was usage by people living in areas where internet connections were slow or expensive.
    Their problem is that they basically promoted feature-crippled and slow stuff as default way to make apps for their platform. This has been further compounded by underpowered devices. And in areas where internet connection is slow and expensive there could also be problem with power as well. Should I say web apps are grossly inefficient in data processing, and therefore woefully power inefficient as well? The more your cpu runs, the more battery is drained, as simple as that. Even Java is rather troublesome in this regard. Though Google gone hell a long way in this regard. Now they even compile it into native ELF executables on program install, if I got it right. Still, these 6 amp-hour monsters imply it wasn't terribly successful.

    It's basically a native application, but written in JS with a sandboxed set of APIs available.
    I see some people don't even know what native is. Native application is when there is native code CPU could readily run - so no heavyweight interpretation or JIT phase takes place. Needless to say, direct execution on CPU with very little fuss attached is very useful for both speed and power consumption. FF OS "native" apps don't have this property by any means. It same crippled inefficient thing and it kinda hard to do something reasonable about it, even when it hurts. Yet Mozilla smartass management preferred to "release now, pr now, think about issues later" approach. Eh, sure, but users had other things around to compare their user experience, at which point Mozilla had very little chance. Even JITed java far more featured and lightweight compared to all that "native" cruft. Apple preferred even more native appls most of time.

    And offline-first web applications can be very resource efficient, but it depends upon how your application is written.
    No way.
    1) JS never meant to be a high-speed solution that is anyhow easy to compile into native code.
    2) HTML itself is fairly featured and complicated thing, its parsing is a rather challenging task on its own.
    3) Typically web things use lame and inefficient data formats that are slow to parse. Oh sure some gamedevs got fed up and requested more efficient approach in at least some places like typed arrays, but seems they finally gave up on this and mostly hold breath for webassembly. Though later is also half-way there.
    4) There're plenty of layers on the way and web tech engines are anything but lightweight.

    Say, Google managed to compile their java things to native ELF executables. And these have no inherent reasons to be terribly slow or inefficient on their own. Good luck to do something like this with web app..

    todo manager in HTML5 with vanilla JS in a few hundred lines of code, and it barely uses any system resources to run.
    ...except truckload of RAM, 10x more cpu cycles (==battery life) and so on. So it "doesn't uses resources" on top-notch dev workstation. But when it comes to battery powered thing, especially with crappy CPU, modest RAM and small battery things suddenly change and you eventually get idea you haven't even launched anything significant, yet device already struggles and even ancient Nokia Symbian brick from prehistoric ages was able to run more tasks, showcase some realtime games with reasonable GFX, and so on....

    consume 150MB of RAM just to display a short grocery list.
    Webmonkeys really prefer to do it that way. This further screwed Mozilla up. Sure, someone of Fabrice Bellard's magnitude could make even reasonably efficient emulator of foreign CPU arch, even in JS! But most webmonkeys are nowhere close. And so it gets like this: thousands of third-rate crapplications that would give whole world of trouble - and good luck to find some masterpiece in all that huge garbage bin.

    Net result is that Mozilla spread efforts and ended up with plenty of third-rate uncompetitive stuff, chasing butterflies all the time. I also don't exactly get why they were so eager to kill everything ppl considered advantage, be it XUL api allowing extensions to deeply change browser behavior or custiomizable UI. And so they preferred to be a third-rate clone of chrome. Who needs third-rate clone if you can just download and use original?
    Last edited by SystemCrasher; 03 September 2020, 09:11 PM.

    Leave a comment:


  • Michael_S
    replied
    Originally posted by pal666 View Post
    they could bundle android, it's opensource for a reason
    How many successful Android devices that don't bundle the Google Play Store or an equivalent proprietary alternative like Amazon FireOS have there been? A few thousand Replicant users? A few ten thousand Cyanogenmod/LineageOS users that choose not to include the Google Android Apps?

    The whole point of FirefoxOS was to have an open source mobile operating system where one vendor doesn't effectively control the whole platform. In that, the project has been more successful than Replicant and Cyanogenmod/LineageOS, but still not successful enough to have a serious impact the mobile operating system ecosystem. Though from what I've read, the FirefoxOS fork KaiOS is used by millions.

    Originally posted by SystemCrasher View Post
    And reason 4: you can have awful user experience with browser-based apps even on flagman devices, not to mention cheap versions. By any means web techs weren't really meant to be high-speed, low-latency, fast-responding things akin to local apps, historically being limited mostly by network latency hiding overall inefficiency and crappy performance - networks were improving far slower than anything else. Attempts to solve these problems actually compounded problems instead of actually solving it.
    The technologies that Mozilla was pushing forward with FirefoxOS were offline-first applications that used local browser storage APIs. The whole point was usage by people living in areas where internet connections were slow or expensive. You open a web page that has the right metadata to be cached as a FirefoxOS application once, it gets cached on your device, and then you only contact the server again to do things like send email or similar. If your application doesn't need an internet connection to function, you use the cached version for the life of your device. It's basically a native application, but written in JS with a sandboxed set of APIs available.

    Browser-based apps on smart phones today are rarely written in that style. They usually require a constant high speed network connection to work properly (mostly for the collection of private data and ad delivery), and that makes the user experience inferior to native applications.

    And offline-first web applications can be very resource efficient, but it depends upon how your application is written. You could put together an offline-first todo manager in HTML5 with vanilla JS in a few hundred lines of code, and it barely uses any system resources to run. Or you could build one using the latest set of tools and have 600 NPM dependencies and end up with 1.3MB of minified JS in your web page, and consume 150MB of RAM just to display a short grocery list.

    Leave a comment:


  • SystemCrasher
    replied
    Originally posted by Michael_S View Post
    You missed reason 3:
    And reason 4: you can have awful user experience with browser-based apps even on flagman devices, not to mention cheap versions. By any means web techs weren't really meant to be high-speed, low-latency, fast-responding things akin to local apps, historically being limited mostly by network latency hiding overall inefficiency and crappy performance - networks were improving far slower than anything else. Attempts to solve these problems actually compounded problems instead of actually solving it.

    And so now flagman devices suddenly have dozens of funny probs:
    1) Thermal management. CPU throttles and can't keep up. Damn bricks literally burn hands even if you manage to dissipate that reasonably, because device is of limited size.
    2) Inefficient crapplications demand for powerhouse instead of CPU. Sure it drains battery like mad. So no even day of battery life?!
    3) In responce to 2) your battery now is quite a large brick as well.
    4) Oh, seems charging that 6 amp-hour brick in a time also quite a challenge.

    So overall ... there is ton of overengineering, and whole thing is getting rather scary, fragile and spitting numerous unpleasant surprises all around. Users obviously curse such user experience and developers alltogether.
    Last edited by SystemCrasher; 01 September 2020, 11:14 PM.

    Leave a comment:


  • SystemCrasher
    replied
    Seems Mozilla pretty much on the road taken by Opera Software recently. That's what you get for hiring bunch of nuts as management, resulting in lack of focus on any particular product - and then hiring cheap and noisy hype-driven monkeys instead of heavy-lifting coders. So their oldschoolers apparently got frustrated bunch of noisy monkeys steals all the fame while deserves it least and gone somewhere else.

    So let's bet when Mozilla would get sold to some chinese company or someone equally shady as well? Seems they don't really have devs capable of developing browser anymore. At most they can afford lame UI shuffling which has only caused fallout and annoyed more users to change to Chrome or derivatives. Servo also took role of snake oil, these got fired as well AFAIK.

    p.s. how the hell mentioned stuff relates to browser development? Even if this "exciting" thing reaches production quality it would be secondary feature, most users wouln't even care if it there at all.

    So a couple of predictions:
    1) Beast is on its deathbed and would be dead in few years.
    2) When it is, ravens would eat their faces, and I'm pretty sure someone like M$ would be more than happy to swallow e.g. rust's Crate - turning them into basically unpaid working force where they exclusively set policies. Just like they recently did on GH - "you've changed browser version, blah-doh, please enter your pincode for the sake of security" [not to mention we collected your phone number] (c) some fellow dev (apparently unhappy about it).

    Leave a comment:


  • Vistaus
    replied
    Originally posted by pal666 View Post
    maybe because nobody told them that "how apple does it" is the right thing
    I dislike iOS almost as much as Android, so I don't think "how Apple does it is the right thing".

    Leave a comment:


  • pal666
    replied
    Originally posted by Michael_S View Post
    But users want WhatsApp, TikTok, etc... and if they don't make PWAs, there isn't much that FirefoxOS developers can do.
    they could bundle android, it's opensource for a reason

    Leave a comment:


  • pal666
    replied
    Originally posted by Vistaus View Post
    The manufacturer of my phone did that, so it's not like I'm using something ported by MrTheBestAndroidHacker or something.
    i was not trying to accuse you. you see, manufacturer did porting, so now we don't know who is responsible for bad outcome - google or manufacturer. or actually we do know: manufacturer sold it to you and is fully responsible for your satisfaction
    Originally posted by Vistaus View Post
    But even on my dad's Pixel 4a, I still find the experience confusing and annoying. But I'll shut up, 'cause apparently I'm not allowed to dislike Android on this forum, judging by your comment as well as a few others.
    it's a subjective matter, i've just pointed out that majority of people don't share your view. maybe because nobody told them that "how apple does it" is the right thing

    Leave a comment:


  • Vistaus
    replied
    Originally posted by pal666 View Post
    i actually should've said "nexus". stock version has to be ported to your device, so someone had to do that with unknown quality
    The manufacturer of my phone did that, so it's not like I'm using something ported by MrTheBestAndroidHacker or something. But even on my dad's Pixel 4a, I still find the experience confusing and annoying. But I'll shut up, 'cause apparently I'm not allowed to dislike Android on this forum, judging by your comment as well as a few others.

    Leave a comment:


  • pal666
    replied
    Originally posted by Vistaus View Post
    I'm not using a Google Pixel, but it's a stock version of Android.
    i actually should've said "nexus". stock version has to be ported to your device, so someone had to do that with unknown quality

    Leave a comment:


  • Michael_S
    replied
    Originally posted by RussianNeuroMancer View Post
    FirefoxOS was ahead of it's time for two reasons:

    1. Even today PWA is not widespread enough while FirefoxOS needed PWA to succeed.
    2. Gecko doesn't have any chances on mobile back then due to old threading model and outdated rendering engine.

    I not sure if FIrefoxOS succeed today because even now when big name PWAs (such as Telegram Web, Uber, Skype, Tinder, Instagram, Twitter, Spotify and even Mirosoft Office PWA!) exist there is no such thing as WhatsApp PWA or TikTok PWA and WebRender is still not enabled by default.
    You missed reason 3:

    Very cheap smart phone hardware when FirefoxOS was launched had 128-256MB of RAM. Very cheap hardware today has 256-512MB of RAM. In a few years, the cheapest smart phone hardware will have 1GB of RAM. At that point, a browser-based operating system with PWAs - and the newer Firefox threading model and rendering engine - becomes feasible.

    And I also agree with your obstacles to its success even now. The WebRender obstacle will be overcome with time. But users want WhatsApp, TikTok, etc... and if they don't make PWAs, there isn't much that FirefoxOS developers can do.

    Leave a comment:

Working...
X