Discussions

a moment ago, cbae wrote

*snip*

The "web" depends on HTTP, but it doesn't necessarily have to depend on HTML.

*snip*

The "machine learning" part isn't web-based. I can pretty confidently say that that code probably doesn't run in a browser. It just so happens what they learn from are resources available on the web, and I'm pretty sure that HTML isn't the only source material their "machine learning" programs work with.

Sure you can parse and learn from anything, but being in a non-complicated standard file format helps. I'll tell you getting data in the right format (feature extraction) is the majority of the work in my opinion. I'm sorry but bytecode compiled against some huge proprietary library doesn't really fit that definition of "data in the right format".

Also I don't think there are well-defined microformats or anything like RDF in XAML. Microformats for sure are used a whole lot on the web these days to convey semantic structure. This was never even a thought when Microsoft developed WPF or Silverlight (R.I.P.) because their focus was to develop a fancy way to put pixels on the screen..

56 minutes ago, evildictait​or wrote

*snip*

Artificial intelligence is an interesting subject, but its progress is entirely unrelated to that of HTML. Artificial intelligence is no better able to to read a website because it is written in HTML than it is able to parse a Silverlight form written in XAML. Indeed, I would contest that XAML (which allows first-class abstractions, such as a date-time-control to be embedded directly in the code) is more semantically rich than HTML which is mainly a collection of DIVs and SPANs with no other context, making it hard to relate a date-time-control on one page as being semantically similar to a date-time-control on another page.

And frankly I would go so far as to say that there are many classes of problems that have been made worse by adding artificial intelligence to them. Clever compilers have led to horrifically hard-to-debug problems because the compiler was trying to second guess what you meant rather than telling you that what you asked for is ungrammatical. In fact, a huge fraction of SQL injections on PHP come from the fact that PHP is so lenient with types that it's happy to allow $someInt = $_GET['foo'] to hold a string value when an attacker submits one, even though the developer and all legitimate users have only thus far passed an integer value. The runtime here is dynamically adapting to unexpected change, but doing so at the expense of the security of the application.

Clippy was hated by many, but was an attempt to add AI to Microsoft Office. Office 2003 also had the collapsible menus which chose which elements to collapse based on your usage, and it turned out that was unhelpful.

AI on the desktop deleting shortcuts that point nowhere has caused anger on this very forum before. Yet another case of the computer trying to be clever and screwing over the end user.

And yet all of this has nothing to do particularly with HTML. Google's search engine isn't written in HTML, and is a huge AI beast. It happily indexes PDFs and DOCX files as well as websites, proving that HTML isn't even necessary to parse content out of stuff online.

HTML is interesting, but you're overselling it. 99% of the tags on the web are DIVs or SPANs which have no semantic content, and 99.99% of all websites don't attempt any AI or have any machine learning whatsoever associated with them.

It's more to with the web and it's nature as a interlinked distributed knowledge store (but a lot of the technology behind the web is explicitly, yes explicitly, designed to make machine learning easier arguably that's the whole point of the semantic web). HTML microformats and new semantic tags in HTML5 certainly help quite a lot in achieving this goal.

It's not a small idea either, when you think about it Google is a web-based machine learning company. We are only scratching the surface here when it comes running machine learning algorithms ontop of the web and already a $300 billion dollar company came out of it..

a moment ago, cbae wrote

*snip*

Microsoft clearly has no clue how to handle [the] Internet.

Generalized further.

Ouch. This might help garner support for the console, but such a drastic reversal also gives people the impression that Microsoft had no clue what they were doing when they first announced the controversial features, so it can reduce confidence in the company.

48 seconds ago, figuerres wrote

*snip*

rest and http are not part of what my question was, they have as far as I can see ZERO to do with the question.

"machine learning" and "AI" are cool subjects but again they are really not part of the question I have.

...

I have a number of apps that I maintain that are point of sale / data entry that gain no benefit I can see from re-writing them to use html to ring up a sale.

Maybe because you aren't asking the right questions? I believe it is rare to find an application that can't be improved, even revolutionized just by adding a little bit of machine learning and location transparency. Elaborate on your POS system.

It's worth noting that you don't NEED web tech to build semantic UIs, nor do you need it do use RESTful architecture. The Web was built to be a globally linked distributed system to represent all possible semantic information (RDFa and increasingly so HTML) and concretely (CSS/JavaScript) which gives it a good structure for input to ML algorithms (esp. ones that need a ton of linked input like deep learning based algorithms); also the collection of stateless globally unique resources concept makes it easy to scale and provide distributed transparency. But you don't need it for anything I described. But it certainly helps..

11 hours ago, figuerres wrote

*snip*

so you have no real answer to the question and you are going to question why I am asking and imply that what I do is not important therefore not worthy of any answers ??

I have put forth a real question that is valid and worth asking and you are telling me not to ask because you have passed some kind of value judgment on any work I do sight unseen?

then you have no interest in actually having a real talk about the question you only want to dismiss me for asking the question.

that makes you as bad as any  Microsoft fan boy who pushes the Microsoft platform.

I gave you a bunch of links. If you want a summary it would be something like "web technology makes machine learning easier" and "web technology was build with location transparency (think: cloud computing) in mind". That's not the whole story of course, learning about the semantic web and how the engineering decisions behind REST/HTTP and helps location transparency and machine learning techniques would help (hence the links).

What I mean is that if your application doesn't use location transparency or machine learning, someone already probably made it 10 years ago in some different form. For me, technically interesting means something that hasn't been seen before - a new technology or software system that solves a previously unsolved problem. It's not impossible for something new and amazing to come along that is simply a bunch of strongly coupled conventional algorithms, but most of that stuff is low hanging fruit and thought of already & might have been sold or used on the open market for decades or more.

9 hours ago, figuerres wrote

*snip*

Bass, I will make one last try here: what fanbaby seemed to have been saying is that using html / css to create the UI of a desktop app is a better way.   I asked him why this was, he said because xaml is dead, I said ok forget xaml I can write an app with win forms or with win32 etc... why is html the way to go if the app is a desktop app that has no built-in need for a web browser or a web server.

So if you have a real answer please explain why it would really be better to add dependency on an external runtime to render the UI of a bog standard desktop app.  the only thing I see is that I have the UI in HTML if I ever need to port it to a different OS and that if I did the html and css right the layout can be changed via css.

other than that we would using html and css be such a great idea ?

right now there is a big push to do this, a few years back there was the MS push to go with wpf

and I know that there are other markup systems out there.

I want to take a good look at what is going on and have a clear / solid set of technical reasons for the next choice in tools and libraries I make.   I am not going to just follow the crowd off any more cliffs.

so just telling me that a thing can be done is not an answer that helps.

on that basis I can do it with Silverlight - it can be done.

or with flash - it can be done.

why html ?

why not some other way that might not need a browser engine ?

It's not very easy to summarize in a single forum post, but it is the philosophy behind the functioning of the web that makes it a more forward thinking way.

Over the past two decades or so, people have devised some pretty profound abstractions on UI beyond drawing pixels on a screen and accepting HCI input in order to make the "User" in User Interface include users that have yet to be imagined.

If none of this matters to you, than I question if this debate is necessary. But given that, I would also question the technological significance of whatever you'd be building.

14 minutes ago, figuerres wrote

*snip*

the question was and is about a desktop app, not a phone app, not a mobile app, not a tablet app.

if I were looking at a mobile / phone / tablet app that would be a different thing.

 

I don't think Windows 8 is solely a tablet OS, also Chrome OS is obviously a desktop OS. Anyway the point is that you can in fact make desktop UIs with HTML5/CSS and without "a web server should be using a web browser to present the UI of the app".

5 hours ago, figuerres wrote

*snip*

so forget about xaml for now and please explain why a desktop applications that has no need for a web server should be using a web browser to present the UI of the app.

as far as I can see the real world benefits are very limited in that case.

I am not saying that you can't do it or that html is bad.

I am asking for facts on why having an application open a browser and depend on that html rendering engine to do the UI of the app is better than something else?

for example if I create a UI with Windows Forms or WIn32  it will be a native UI with full access to the OS  and the native Windowing system. it will also not be dependent on what browser is installed on the pc and to what happens if the browser is not there etc...

also let's talk about the issues of writing an application and how it will have to bridge to the browser runtime and exchange data with it. I can see a number of cases where the security of the data exchange may be an issue - for example in a medical facility with HIPPA compliance rules and so on ....

please explain ?

False dichotomy. You have an HTML5 based app that can make native calls as long as it is not running in the browser's security context. People have been making mobile apps in this manner with frameworks like PhoneGap for years.

Even Microsoft has been a huge proponent of this, and they have something called WinRT in the latest version of Windows that basically does that..

1 day ago, PopeDai wrote

Also the Office app is missing a lot of functionality that's present in iWork for iOS, it feels very half-baked.

I feel like iWork is a bit of a loss leader for Apple to make their iProducts enterprise friendly.

I feel like this thing is just to sweeten the deal for Office 365 subscribers who probably have Windows PCs with the full Office suite at work.