Re: Emacs design and architecture. How about copy-on-write?

[Dmitry Gutov <dmitry@gutov.dev>]

On 21/09/2023 16:44, Eli Zaretskii wrote:
>> Date: Thu, 21 Sep 2023 16:30:00 +0300
>> Cc: luangruo@yahoo.com, acm@muc.de, incal@dataswamp.org, emacs-devel@gnu.org
>> From: Dmitry Gutov <dmitry@gutov.dev>
>>
>> On 21/09/2023 16:00, Eli Zaretskii wrote:
>>> You are basically proposing us to provide broken, incorrect, and
>>> outdated display in some cases.  That won't fly, so it's a
>>> non-starter.
>>
>> If might prove an okay experience, if the "outdated" part is within
>> certain bounds. Like if the current buffer refreshes at 30/60fps, and
>> the mode-line at 10/20fps, for example.
> 
> That's not what was proposed.  The proposal was to do a
> "quick-and-dirty" layout first, and use the results if the "accurate"
> layout takes too long.

I think I've also commented on the potential problems in my first reply 
to that. But since it's just an abstract suggestion at this point, we 
can try and see the constructive aspects of it.

> The quick-and-dirty layout can only be based
> on ignoring changes in fonts and disregarding images and stuff like
> that, so the screen lines will be wrong, and you will see them jump
> and/or rearrange after some delay, which could be some seconds.  Some
> browsers do that when they need to download images, and it always jars
> me when that happens.  Usually, I also lose the place where I was
> reading.

Some details on what the browsers really do:

- When the images are not downloaded yet, it's possible that the browser 
doesn't know how to lay out the page yet, and the time to download sad 
images is not guaranteed. So what you're describing is really the 
best-effort approach, which has been most noticeable in the past, over 
slow network connections (or in countries/locations with poor network 
coverage even nowadays). Either way, it's a best practice for the web 
page authors to include the dimensions of the images in the <img> tags 
-- when they are present, the layout can be done correctly even while 
the images are not downloaded yet.

- If a web page is complex enough, and I scroll it down after opening 
very quickly, it can happen so Firefox doesn't manage to render the page 
quickly enough, perhaps due to other load. When that happens, it shows 
the bottom of the page in a sort of blurry haze (which is replaced as 
soon as it's ready). Overall, it feels good enough usability-wise, 
because it doesn't look jerky or "blinky" or etc. And of course it's a 
rare occasion overall.