Re: Sluggish user experience on Facebook

David Goldfield

I realize that the tone of my last message may have sounded more negative than I might have liked. I should also say, in fairness, that NVDA, in most cases, is not overly sluggish. The biggest issue I've had was with Word but enabling the experimental UIA support has really made a difference. I'm doing a few comparisons to see if some synthesizers are more responsive than others. Lately, I have been using the legal Eloquence driver from Code Factory but I find that the Vocalizer voices from Code Factory have really improved with UIA support enabled, particularly when typing in a Word document. If we could see a similar change on the Web it would really improve NVDA's responsiveness. Tony Malykh's experimental focus toggle add-on also seems to help with responsiveness.

David Goldfield, Assistive Technology Specialist WWW.David-Goldfield.Com

On 12/28/2018 9:42 PM, David Goldfield wrote:

I have a few comments I'd like to add to this thread.

First, I know that one of the reasons why the latency issues don't always plague me as much is that I usually disable character and word echo while typing. Silent keyboard entry means that I don't encounter latency issues, at least for that particular task.

I also suspect few people are sounding the alarm about this because, if I had to guess, few people may actually be running into serious latency issues. At work I have access to two separate laptops. Both of them have 16 GB of RAM and an I7 processor and I must say, in fairness to NVDA, that the latency issues I run into at home are almost nonexistent on those particular laptops. When compared to JAWS, I still notice a very slight difference but they are not significant enough to get in the way of typing, editing and reading text. This is true even in Word 2013 and Word 2016, where performance is certainly acceptable.

Where I notice the issues we're talking about are on my home Optiplex desktop. It's much slower compared to the laptops I use at work, with 8 GB of RAM and a dual core Athlon processor, 2.8 Ghz. I'll admit that when this PC was first made I believe it may have first come preinstalled with Windows Vista. When I bought it as a refurbished unit it had Windows 7. I added more RAM, an SSD and upgraded to Windows 10. Some might argue that maybe I should consider purchasing a new computer and, eventually, I'm sure that I will. However, when I use JAWS 2019 I notice that the latency I experience with NVDA is practically nonexistent. In no way do I mean to start an NVDA vs. JAWS debate on this list. That's not the purpose of my response. However, one of the things that I think should be a high priority with a screen reader is responsiveness and eliminating latency wherever possible. For most people on most machines NVDA is pretty good in this regard which, again, is probably why so few people are reporting issues. Being a small organization NV Access may have to choose to not prioritize on an issue which may only be affecting a small percentage of their users and this makes sense to me. However, I would respectfully counter that by saying that NV Access's goal is to put a high-quality screen reader into the hands of any blind person who needs one, regardless of their location and of the language that they speak. While many of us either have access to or can afford a new computer the fact is that many cannot and I am sure there are many consumers around the world who may need to use slightly older technology with their screen reader. I realize that we have to strike a reasonable balance here, which is why NVDA, as an example, no longer supports Windows XP. This makes sense to me. However, like Tony I wish that some refactoring could be done with the goal of improving the responsiveness of the speech, with the goal of making the experience more like JAWS, if not better.

Again, this is not meant to start a screen reader debate nor is it meant to trash NVDA in favor of JAWS. I would just like to see these latency issues addressed. However, to be fair a company isn't going to make such a massive development effort a priority if the perception is that this is not a problem for most or even nearly all of its users and I can also respect that.

David Goldfield, Assistive Technology Specialist WWW.David-Goldfield.Com
On 12/28/2018 12:32 PM, Tony Malykh wrote:
Hi Felix,
I'm on your side - I think sluggishness in NVDA is a problem, in fact
a big one, and I wish NVDA developers could do something about that.
Facebook is not the only web site that has these issues, I experience
the same problems on gmail and private web sites of company I work
for. I have to switch to JAWS to browse these sites every time just
because all these symptoms with NVDA are unbearable and makes these
web sites completely unusable.

I think another problem with these issues is that the majority of NVDA
users don't seem to be bothered by this issue, judging from reaction
on this list. I think that:

1. People's typing speed differs a lot. I tend to type very fast,
which probably exacerbates this issue for me. People who type slower
might not be that annoyed by this issue, as for example the focus is
sucked into edit boxes only when you're going through them quickly.
Felix, are you a quick typer as well?
2. Maybe people tend to stay away from these web sites that they
cannot figure out how to use. Since Facebook is glitching with NVDA,
blind people will have hard time to learn it and will use for example
twitter instead. It becomes a chicken and egg problem. NVDA devs won't
fix an issue that doesn't affect many users. And it won't affect many
users because the issue is there to begin with.
3. In my experience the severity of this problem depends on the
browser. Glitching in Google Chrome seems to be several times worse
that in Firefox.

Anyway, these are just some thoughts. I wish I could help convince
NVDA devs to have a closer look into this, but looking at github
issues, this problem has been known since at least 2012 so I think
it's unlikely it's going to be fixed any time soon. Unless we can find
enough people bothered by this and maybe like file a petition. If I
had more time, I might have tried to fix it myself, but unfortunatley
I don't have much spare time. So in the meantime - I'll have to switch
back and forth to Jaws.


On 12/28/18, Felix G. <constantlyvariable@...> wrote:
Hello Enes and list,
sorry for my potentially confusing remark. I was not contrasting our
user experience with that of sighted users. Rather, I was implying
that some peculiarity of Facebook or the affected browsers might give
screen readers in general a hard time, and not just NVDA specifically.
So I did some testing, that is, I performed some comparative screen
reader theology, with the following results:
The late Window-Eyes, digital peace be upon it, occupies the slow end
of the speed scale combined with the high end of the accuracy scale.
You tend to get to wait quite some time for an answer, but when that
answer finally arrives, it is accurate to the degree of fully
supporting the weight of subsequent interactions.
JAWS seems to take quite the opposite approach, erring, when it does,
in favor of speed. There is the occasional unsolicited switch to focus
mode, as well as the occasional double-speaking of text, but the
snappy experience minimizes these frustrations.
NVDA currently takes what I call, from my perspective, the unfortunate
middle ground of this tradeoff. Applying this observation to Facebook,
it means that the experience is slow enough to feel sluggish, and just
unpredictable enough to feel nondeterministic as well. One can clearly
spot all the good intentions with which this particular road to hell
has been paved, so I wouldn't consider this proof that anything is
being consciously ignored, but the result, at least for me, combines
the worst of both worlds enough to make me switch screen readers when
using Facebook.
If, as Enes suggests, this observation applies to a broader category
of websites of which Facebook is just one prominent example, the
approach taken by NVDA in interacting with browsers deserves
reconsideration. Given the dynamic nature of todays's web, I strongly
suggest favoring speed over complete consistency. Selectively
decoupling focus from virtual cursor position, and re-coupling them on
demand, may be an effective first step in the right direction, as some
isolated speed cost when activating an element is allowable while a
permanent speed cost while navigating is clearly not.
All the best,

Am Do., 27. Dez. 2018 um 22:28 Uhr schrieb enes sarıbaş

Yes, as sighted obviously do not have this problem, or we would've heard
more about by now.

On 12/27/2018 7:49 PM, Felix G. wrote:
so you're certain it's an NVDA issue rather than a Facebook or browser

Am Do., 27. Dez. 2018 um 20:43 Uhr schrieb enes sarıbaş

This is an NVDA issue which has existed for about a few centuries,
is also ignored by the devs. This happens not only in facebook, but in
many websites with dynamically changing content, or with dialogues,
instance, the Oxford scholarship online database my university
subscribes to. This has to do with the way NVDA interacts with the
browser, however a better way was not implemented.

On 12/27/2018 7:29 PM, Felix G. wrote:
Hi everyone,
is anyone else noticing that when arrowing or otherwise navigating
Facebook's standard view, the interface tends to slow down to a
In addition, I am experiencing the virtual cursor randomly jumping
back a few lines, resulting in lines being read twice, and random
switches to focus mode.
Are there any known remedies, except of course switching to mobile
view, which is not even recommended to blind users by Facebook
All the best,

Join to automatically receive all group messages.