Wandering Thoughts archives

Some things I believe about importance and web page design

Here are some things that I believe about web page design.

If you have a bright red element on a page that does not otherwise use red as part of its colour scheme, this element should not merely be the most important thing on the page, it should be so important that the reader must look there first and foremost. In at least the west, we're habituated to red as an alert colour, probably the primary alert colour. A solitary red thing is practically screaming at you, jumping up and down and demanding your attention. It should be worth it and really, it had better be.

If you have a bright red element that you lock as a floating element so that it's always present even when the page scrolls, this element should be so important that the reader needs to always pay attention to it. A constantly present red element is a constant alert yelling at you. I actually tend to think that if it is so important, maybe there shouldn't be anything else on the page (or at least no other content).

(In fact, in general any always present element should be very important. Screen space is a precious resource on at least mobile devices (which are an increasing amount of web browsing), so any space you reserve for headers, footers, sidebars, and so on is space that shrinks an already small content area even smaller. People with small screens can get quite irritated about this and yes, they would much rather scroll up to get back to your navigation than have it take up a quarter or a fifth of their screen all the time.)

If you have or are tempted to have a red element on your pages, especially an always visible one, I very strongly think that you should be considering whether it really is that important. Is it an alert that visitors should be paying some amount of attention to all the time, to the degree that some of them may find the result unreadable? Is it more important than the actual content of the web page? Or is it perhaps somewhat less important than that, and so maybe it should not be red, or always visible, or especially not both at once.

(My personal view is that there is very little information that is that important. Your web pages are hopefully about your content, after all, at least for normal websites. About the only case for such an all-consuming alert that I can come up with is content that is actually now dangerous and is being retained only for historical reasons; here, people not reading the content is a feature instead of a bug.)

In not unrelated news, the PowerDNS docs web pages for the 3.x release currently look like this [png]; 3.x is not the latest release, but it is one that is still very much out in the field (it's in Ubuntu 14.04 LTS and Fedora 23, for example). I personally do not consider 'you are reading the 3.x documentation, here is the latest documentation' to be the most important and pretty much all-consuming thing on the page, but it's not my website.

(Nor will I be even attempting to send in a patch, because the only patch I would write is one that deleted the entire 'this is older documentation' alert. There is no possible way of fixing this within anything that looks and acts like the current setup, and I have to assume that the people who created the alert feel that it is really that important. In that case we have a disagreement not about presentation, which can be patched with HTML and CSS, but about web page information architecture.)

I've started using the Firefox addon Self-Destructing Cookies for some stuff

I previously wrote about two models for dealing with cookies in Firefox and how I was someone who followed the 'never accept' model while Self-Destructing Cookies followed the 'accept then remove' model. Well, as it happens I've recently started using SDC in some of my Firefox profiles, for what I think of as a good reason. Namely, it's really easy to get started with SDC.

My main browser is not logged in to sites, in addition to basically not accepting cookies. For a few sites I use regularly or want convenient fast access to, this is inconvenient enough that I maintain a separate Firefox profile; this profile is set up to retain cookies so that I stay logged in. I recently realized that the simple setup of these profiles meant that I was retaining all cookies, not just the cookies for the site I care about. When I actually looked, I found that I'd built up quite a collection of tracker cookies that I really didn't like.

I could have reacted to this by setting up a strong anti-cookie addon in those profiles and adopting a default-deny policy. But while I can clearly identify what site I have to allow for my login cookies to work, some of these places may need side cookies and the like. I decided that I didn't really have the energy to tackle this; I wanted a simple solution that was easy to enact and almost surefire. This is a situation that Self-Destructing Cookies is ideal for. I could install it, tell it to permanently keep cookies from the main site, and be done.

So far I'm quite happy with using SDC this way. It's been a no fuss, no muss way of choking off a great deal of cookie abuse at minimal cost in time and attention.

(As with almost all Firefox addons, I've found that I have to tell it to be quiet about notifications. For some reason, many addons assume that I'm eager to hear about everything they're doing and blocking for me. This is very much not the case. I'm perfectly well aware that the web is a horrorshow of trackers and privacy abuse, and I don't care about the details of who is abusing me this time.)

(Because this is for profiles that I don't keep running all the time, I don't know if SDC leaks memory in normal ongoing use, as all too many addons do.)

Firefox addons seem unfortunately prone to memory leaks

As is now traditional, I'll start with my tweets:

I look forward to a hypothetical day when it's difficult for Firefox extensions to leak memory like sieves. That day is not today.
It's depressing that at least half the time I audition new Firefox extensions, they immediately cause my session to bloat up madly.

It's like clockwork by now; I'll decide to try out a new extension (this time around I believe it was Cookie Monster), and at most a few hours later I'll see that my Firefox has monstered out into high and steadily growing memory usage. So much for that extension.

I have to assume that this is not a universal experience with Firefox addons, especially popular ones like Greasemonkey and Stylish. Clearly there's something unusual about either how I use Firefox or how my collection of addons interact with each other that makes me especially prone to it. I don't think it's just leaving Firefox running with pages open, because I'm sure that at least some people these days leave tabs permanently open on GMail or Twitter or Slack or Facebook or any number of other 'constant on' sites. Perhaps it's my use of windows over tabs, or combining multiple windows with some having tabs, but even that feels stretched.

(I don't think it's using Firefox Nightly either; I've tried out extensions in a stock Firefox and observed bloat symptoms there if I push it.)

It would be nice if Firefox Electrolysis would fix this, because then I'd really get something for the problems it's going to bring. But I don't actually believe that that's going to happen, and in fact I expect e10s to make my situation worse, because it'll undoubtedly cause me to have to change some extensions (and thus hunt for ones that not only work but don't leak).

(It's possible that e10s could help here. Given the pervasiveness of addon memory leaks, there are clearly Firefox addon APIs that are hard to use without leaking memory. If e10s replaced them with more leak-free APIs, we could win. But I don't really expect this because I don't think this is a core part of Electrolysis's purpose.)

Incidentally, this is one reason I am so grumpy when Firefox does something that breaks one or more of my extensions. My current set of extensions is extremely carefully chosen to not just work but also to not leak memory. Finding replacements with both attributes is both a bunch of work and also not in any way guaranteed. I've certainly had to drop addons in the past when they leaked and there was no non-leaking replacement.

(I suppose for big addons like Greasemonkey I should try to find their bug reporting system and report the issue, just on the odd chance that this gets things fixed. But making bug reports is exhausting.)