Hacker News Reader

Comment by fouc | Hacker News Reader

What if the browser had visible page score next to the search field?

It could be a score based on asset sizes, memory impact, cpu impact, etc.

By making it visible, people would be more cognizant of which sites have poor experiences or have a big impact on their computer.

View on HN

By making it visible, people would be more cognizant of which sites have poor experiences or have a big impact on their computer.

You can't simplify this problem down to "This uses more RAM than an arbitrary threshold therefore it's a problem." If I spend 99% of my time using an app then I want it to cache hundreds of megabytes of data in to memory so I can work fast. Saying it's a bad experience if it does that is wrong.

That's not about right or wrong, and by the way, the proposed indicator indicates numbers, not wrongness level.

I'd add to it the slider "cache this app more-less", because you know how do you usually use this app or page.

And the "complain" button with auto-redirect to some support, or uservoice form of this app, if there is any, when e.g. I moved the slider to the "less" as possible and even then it still eats 200 megabytes of RAM.

Take a particularly bad example: Gmail uses over 150MB of RAM, and achieves similar results to what FastMail does with 11MB. The main difference is that Gmail includes the Hangouts chat system—and indeed, last time I checked Hangouts was responsible for most of the megabytes downloaded, most of the time spent and most of the memory consumed. Suppose then we decide to let Hangouts off scot-free for its profligacy, calling it 50% of the footprint, and just focus on comparing the email part of the system.

Gmail is still using an estimated 80MB for what FastMail needs only 11MB for.

Where is this 80MB coming from? Is it caching 80MB of emails? I think not. You’ll be lucky if that 80MB corresponds to even half a megabyte of email.

Most high memory usage is not because it’s caching things so it can work fast. Most high memory usage is simply because the app is inefficient.

I'd be interested to see how the offline capability of Gmail ties into this. Whatever it uses, localstorage or service workers or whatever, it does a pretty dang good job.

I can’t speak about how Gmail’s offline functionality works, but I would expect it to be broadly similar to the basic approach we’re planning on in FastMail, which is that you start by simply adding a persistent cache. There’s already a robust object syncing protocol in place (JMAP), and you already have what amounts to simple tables (Mailbox, Email, CalendarEvent, &c.), so all that’s really changing is that you’re moving most of that cache onto disk rather than keeping it in memory. Most likely, content is moved out of the JS heap into an IndexedDB. I’m not sure what the memory characteristics of IndexedDB are yet, but at least theoretically that should be memory that can easily be freed, because it’s all flushed to disk. There’s the possibility that the browser may keep more in memory than is necessary, but I would actually expect memory to be a little lower in the presence of service workers, rather than higher.

I'm not convinced that such a number could be calculated in such a way that wouldn't be utterly meaningless to all but the biggest nerds.

The resources a 'web app' 'should' use is highly context dependent. As a web developer, I can determine some of that context, as I know what functionality is resource intensive. I don't think that you can distill that down in any useful way.

> such a number could be calculated

Compute this from the start of page load. Adjust the scale of the final output if desired.

    # automatically scaled to local hardware
    # in units the user actually experiences
    t := "total wall-clock CPU time used (in ms)"

    # yes, allocations - NOT total usage
    m := "Total heap allocations (in kB/kiB)"

    # magnitude of about:blank
    a := typical_m_for("about:blank") *
         typical_t_for("about:blank")

    # score is computed similar to decibel
    score := 10 * log10( (m * t) / a )

A variant of the score that excludes media data in <audio> or <video> tags should also be computed. Both versions should be presented.

Continuously update these numbers, so the user can see the impact of any background JS/etc.

> utterly meaningless to all but the biggest nerds

Please don't assume people are stupid. If you give people real data consistently and it affects their life, they will figure out how to use it. Their interpretation may not be technically rigorous, but it will mean something to them. Also, most people will understand and expect the score for youtube to be (much) larger than a messaging service like twitter and both larger than a simple static text-only page.

> as I know what functionality is resource intensive

You only know this an isolated abstract sense. You do NOT know how much network bandwidth, CPU time, or RAM the user actually has available and wishes to use for you page. They may be using their bandwidth for other important things on another computer. Their RAM an CPU might be needed for other uses - inside or outside the browser.

Approximately nobody only uses one application at a time. Your app or webpage is almost always going to be competing for resources in an environment you do not control and can never truly understand (you do not know the user's ultimate goals, environment, or requirements/restrictions).

I'm wondering if this could be made a relative score instead of an absolute one that measures the current site's memory consumption relative to your available free memory. It might be more useful to see the impact it has for you.

That seems like a convoluted scheme to get "normal" people to care about something that you care a lot about, and they don't.

They don't care because it's not a problem for them. Yeah, maybe they could have bought a cheaper phone if everyone had spend twice as much time writing the code. But hardware advances have actually caught up with requirements (and then some), and all these sites work fine on even lower-end current phones.

You care about it like a watchmaker cares about the mechanical drive of his watch.

They don't need a watch. They have smartphone. It's 8 magnitudes more precise than your mechanical watch.

It is absolutely a problem for them. It is what makes them pissed off and ruining my day by getting in anger fits over how their phone is shit (it's not their phone - it's the accumulated bloat of the apps they use). It's what makes them buy a new computer ("my old one was too slow").

It's just regular users know so little about technology that they accept what they're given without any question. "Thinks work slow so it's definitely the fault of my computer, maybe it has viruses".

I was about to say this. Computers are not getting slower over time, the programs get worse over time.

I care about bandwidth because I don't have unlimited data on my mobile plan.

I wish I knew how much each site/domain was costing me. Then I could have a better idea which sites are profligate wastrels to be avoided and which sites I can feel free to visit at will.

I agree, but I believe this is referring to memory usage, not data transfer/usage over the network. But that number would be interesting if more readily available.

I use the uBlock origin badge in my browser as a proxy for making page judgements - I avoid revisiting sites with a high number (>5?) of blocked elements...

Reply on news.ycombinator.com