Am 04.12.2015 um 16:19 schrieb Ivan Vučica <[email protected]>:

> Here's a proper response now that I am at a computer and have found the 
> original mail to respond to :)
> 
> Side note: we broke the 100 posts mark in the thread -- congratulations, 
> everyone, on a centithread ;-)
> 
> On Fri, Dec 4, 2015 at 6:52 AM H. Nikolaus Schaller <[email protected]> 
> wrote:
> Am 03.12.2015 um 20:57 schrieb Ivan Vučica <[email protected]>:
>> 
>> We have an HTML viewer, and it's quite good at that. But it's not a browser, 
>> and should not be -- it's too much work for too little benefit, not to 
>> mention performance can only suffer.
> 
> Well, with this way of argumentation you could say:
> 
> GNUstep is an old-fashioned technology demo and is quite good at that. But it 
> is not an OS or desktop and should not be -- it's too much work for too 
> little benefit, not to mention performance can only suffer.
> 
> I.e. such arguments do not help.
> 
> You are misinterpreting my words.
> 
> I did not run SWK and Vespucci locally (never got around to it), but Riccardo 
> has demoed it at the Dublin meeting. I am impressed by it, but not as a 
> replacement for a full browser. Instead I view it as an excellent lightweight 
> tool for viewing simple HTML documents, such as our documentation. I value it 
> as GNUstep's graphical alternative to w3m, links and lynx -- not as a GNUstep 
> alternative to Firefox and Chrome.
> 
> Does that clarify?

yes.

But... you rate it on the status it did reach by teh last commit ~2 years ago 
and not by what could be achieved by more actively developing it.

Anyways I think the arguments have been exchanged and since a community project 
does not making centralized decisions but everybody can decide on her/his own 
to ignore it or to work with it,
we IMHO do not even need to come to a conclusion..

BR,
Nikolaus


>  
> 
>> 
>> 
>> - Can it do 3d transforms?
> 
> If someone writes a patch (and the backend supports)
>  
> 
>> - Does it do WebRTC and WebGL?
> 
> 
> If someone writes a patch (and the backend supports)
> 
>> - Will I be able to access Outlook.com and Google Docs in it?
> 
> If someone writes a patch (and the backend supports)
> 
>> - How well do web applications self-declare support for it?
> 
> If someone writes a patch (and the backend supports)
> 
> But this is exactly the point. How much 'someone should write the patch' does 
> one go into to support modern web? How many developers work on WebKit, Blink 
> and Gecko?

It could be hundreds as long as we more or less tell the first one to join that 
there are no others and he better does not join.

> 
> SWK has value and is worth improving. But a GNUstep-based desktop needs a 
> modernized engine wrapped inside a GNUstep-based chrome*, with bitmaps for 
> certain commonly-themed form elements are provided by GNUstep's theming code. 
> Which is where Vespucci-over-Blink using CEF** becomes interesting.
> 
> * chrome, in this context, being the term used for 'web browser's elements 
> displayed around the contents of the page'.
> ** https://bitbucket.org/chromiumembedded/cef
>  
> 
>> 
>> There are really good engines that support running web-based applications 
>> well.
> 
> Yes, they have become really big beasts to support thousands of pages of 
> standards.
> 
> Yes, which SimpleWebKit should not be, but which 
>  
> 
>> Yet even long-standing, reasonably well written engines such as Opera's 
>> Presto are being dropped.
> 
>> 
>> SWK fills a need for a performant HTML viewer, but is not a proper web 
>> browser engine.
> 
> As an observation of the current status you are very very right. But is it 
> limited by principle or by manpower?
> 
> Manpower.
> 
> But I don't object to viewing that as a matter of principle either.
> 
> Once you start doing all the crazy things modern browsers are doing, how 
> simple is SimpleWebKit going to be

it is objective C only and does not do any non-straightforward tricks.

> and how fast is it going to be?

Depends on optimisation.

> Will SWK, for security reasons, start separating tabs into different 
> processes?

If GUI and Base can support this.

> Will its renderer run in separate process(es)?

If GUI and Base can support this.

> Will its JS implementation start supporting multiple security contexts 
> (which, I recently discovered, Chrome does -- extension code injected into a 
> web page seems to run with different permissions, even though it lives in the 
> same web page)?
> 
> It's not a problem of can it eventually become a beast supporting thousands 
> of pages of standards. 
> 
> As its principal author, do you personally view that it should?

It should support the most important standards.

>  
> 
>> 
>> A good GNUstep browser would use an existing engine, but integrate with a 
>> GS-centric environment:
>> - by using GNUstep's theme for its chrome,
> 
> Isn't that working out of the box? Vespucci & SWK just use the NSView 
> subclasses provided by GUI.
> 
> It does.
> 
> But please consider what a modern-day user expects of something that s/he 
> would call a browser: if we assume that on top of that list is "it should 
> open Facebook and Gmail" (arbitrarily chosen websites), then we do not 
> currently have a fully-featured GNUstep browser.
> 
> Let's look at options:
> 
> - Vespucci, being a GS-based chrome for SWK, definitely is themed, that is 
> not disputed. But it cannot open Facebook or Gmail.
> - Chrome, being a GTK-based chrome for Blink, is fully featured, but is 
> unaware of GNUstep's themes, save panels, et al.
> - Firefox, being a GTK-based chrome for Gecko, is fully featured, but is 
> unaware of GNUstep's themes, save panels, et al.
>  
> 
>> - by exposing GNUstep's Services in its textboxes and for its images,
>> - by using GNUstep's save panels, by understanding the concept of bundles, 
> 
> what do you mean/expect by that?
> 
> If I right click->"save as" on a photo, whatever is generally accepted as "a 
> GNUstep browser" should display GNUstep's save panel.
>  
> 
>> - by storing its preferences and cache inside GNUstep's folder structure 
>> (~/GNUstep/), 
> 
> AFAIK it uses NSUserDefaults and WebPreferences which can be adapted to 
> GNUstep's folder structure (if they don't do already).
> 
> Not disputed, if we talk about Vespucci + SWK.
> 
> But will it *currently* open and correctly render cnn.com (and any 
> interactive elements it may have)? Will it do so in the next two years? If it 
> ever does, it will become as bloated and slow as any other engine. Should it?
> 
> If we talk about Chrome and Firefox, neither currently does that.
>  
> 
>> - by registering web shortcuts (e.g. .url files) with GNUstep's extension 
>> registry, 
>> - by using GS menus (whatever they are as configured by the user) and 
>> therefore by using GS-like keyboard shortcuts
> 
> What is missing there?
> 
> In Vespucci, nothing is missing there.
>  
> 
>> - in case we have a 'quit app quickly, but restore NSDocuments and its 
>> windows on start', integrate with that
>> etc. 
> 
> Do we have that?
> 
> Only Vespucci (which will not serve as my browser any time soon) could 
> hypothetically integrate with such an upgrade to NSDocumentController.
>  
>> 
>> Providing an alternate implementation for use by Vespucci seems useful.
> 
> You can extend Vespucci and replace SWK if you like. It should in theory be 
> as simple as replacing the WebKit.framework or linker search path.
> 
> Which is exactly the ideal outcome.
>  
> 
> But I don't want to argue at all against any alternatives to SWK and 
> Vespucci. I just make aware that "something" exists.
> The alternatives may be much better and easier to develop, but do not exist.
> 
> I am aware it exists, and I am impressed by SWK. 
> 
> But it does not, and should not, serve every use.
>  
> If we would contribute as much code as we recently started to write e-mails 
> what *should* be done, there would be more progress :)
> 
> I do not disagree here.
> 
> Of course, I see other areas in GS as currently needing more love than 
> wrapping a browser engine in a way that can be integrated into Vespucci. :-) 
> For example, I would really like to be able to direct people to 'apt-get 
> install gnustep-session',

me too since I can do that for QuantumSTEP for years (I still struggle a little 
by libobjc versions in Debian 7.x vs. 8.x)...

> or 'download the live CD ISO here'. Crucially, I would also like to be able 
> to provide an updated gnustep-session with a single command.
> 
> Simple goals, but they are proving to be tricky to actually bring to 
> completion.
> 
> If achieved, perhaps I will still feel that the lack of a browser integrated 
> into the gnustep-session impacts the UX. But I'm not there yet. :-)

That is where I completely agree - reliable and fast installation is more 
urgent than a good browser.

BR,
Nikolaus


_______________________________________________
Discuss-gnustep mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Reply via email to