Gracefully Dropping Support for Dinosaur Browsers in Trello | Hacker News Reader

Gracefully Dropping Support for Dinosaur Browsers in Trello

blog.fogcreek.com

59 pointsjoeyespo52 comments

I just got a report from NASA's systems admins saying they have approved IE9 for general rollout to the work force. The key paragraph:

"Beginning Tuesday, March 25, Internet Explorer (IE) 9 will be deployed to all ACES Windows computers. Not only does IE9 improve the security posture and comply with federal requirements, but it also has a new look and many new features. While there are later versions of IE available, the EUSO and HPES must do due diligence in analyzing each version from a security and compatibility perspective to ensure the protection of NASA systems. The ACES team will continue to evaluate and roll out the latest and safest versions possible."

So I had to laugh when the article stated, "Some of the browsers we supported at launch just can’t be considered modern by today’s standards. I’m talking about Internet Explorer 9 specifically."

Our largest clients (both large banks over here) are still on XP+IE8 on their standard desktops. At least we are no refusing to support anything older (there are still pockets of IE7 at RBS it would seem) but I'd love to refuse to support IE8 too...

You should start refusing on April 8th, at which point Windows XP is unsupported and insecure.

I'm betting it will be just as safe/unsafe on April 8th as it is on April 7th. Yes people should have gotten off XP by now, but real life isn't always perfect.

There will not be a massive XP infection on April 8th, it will come at least a few weeks after that ;)

It might very well come on April 8th, since malware authors who find exploits in the period leading up to that date would wisely wait until right after the no-more-patching cliff to release their new 0days.

Considering how many Firefox flaws were saved up for the recent hackathon (since the rewards are 6x what Mozilla usually pays as a bounty), I'm betting on at least a couple critical ones being unleashed the day after, though it may take some time for them to be caught.

We couldn't. We don't sell off-the-shelf, we are selling on contract and that contract states that we will support their default build. We could argue the point at the next contract negotiations but if we simply refuse they will simply take their business elsewhere. Our concerns have been raised a number of times and will be raised again and minuted each time, so if there is a problem that comes down to their use of an outdated browser we won't (ligitimately) get it in the neck.

We are charging extra for any custom work to allow for testing and bug resolution specific to ancient combinations, and refusing to address superficial issues at all without a full change request and appropriate payment, but we simply can't refuse to aupport XP+IE8 without telling them to take their £20,000/month elsewhere.

We do have big new potential customers on the horizon, and they are more up-to-date in that respect. Maybe if these people don't upgrade soon and we do get huge new contracts, we'll be in a position to say "your choice, upgrade or go be a problem to our competitors" come the next contract round. I hope so.

I was just pitching a client that was on IE 8. D'oh! Very hard to keep things working on older stuff. That said, it seems very surprising that overall usage of IE was so low. I would have thought it to be at least 20%.

Trello launched from day zero, in Sep 2011, without IE8 support. At that point, IE <= 8 was about 80% of the IE crowd[1]. FogCreek said that IE8 support was coming, but it never materialized, and they eventually just passed on.

So the truth is that Trello always stayed pretty much ahead of the IE usage curves, so that most people with IE willing to use it, simply couldn't.

I guess this is the reason why they don't have many IE users in the first place.

[1] according to W3Counters, which is the first I found with 2011 numbers readily available.

It’s true that Trello never supported IE8, but we never promised or intended to add IE8 support. We wanted to support the browsers of today and tomorrow, which was IE9 at launch.

I stand corrected then, you never even tried to get the IE crowd.

But my point that you're misreading numbers remains. You see low IE9 usage because you don't have IE users, because, in the previous years, up to 80% of them couldn't use your product.

And now you're using these numbers to back your decision that you don't need IE9. Guess what? You won't see IE adoption grow much over the next years as well.

IE users means enterprises. By dropping IE9 now, you're again running faster than them. That's perfectly fine and I'm actually jealous, I wish I could do the same. At the same time, I'm sure it has a measurable impact on the potential market of your Trello business class product.

This makes a lot of sense. If you don't ever support a browser, then there's no apology required when you remove support of it.

Yeah, we just recently got our IE upgraded to 9 on our Army (MAMC) network. FF 24 ESR too... moving on up.

Unfortunately many networks on many of the bases are still using IE 7 and 8, so we have to support those some of the time (project dependent).

Active users at work: 38% IE7, 8% IE8, and 45% IE9

I still remember the backlash from when they were going to require IE10 for VS2013.

I am starting to wonder if the overhead of doing an API first design is that much more than integrating your domain model into the web app directly. The benefits of having an API backend really seem to shine in the scenario that Joel is describing.

Branch a version of the codebase for the older browser, remove all the support for old browsers from trunk and off you go.

I think the main decision that Joel has made is that they will not shoehorn new features into the older browsers and instead force them to upgrade if they want the new features. May not work for other web apps but seems to be a good compromise.

Are there any other known examples of websites that follow this API first design? I would love to see more discussion about this design choice.

I think the biggest example would be Amazon. The biggest advantage I see is that Joel's approach still allows some level of iteration on the domain model and all the associated business logic.

It seems that mobile is pushing this type of design as it makes the most sense (so Facebook, Twitter and Google follow this approach quite a bit). It looks like the logical next step is why can't the desktop browsers be just another "View" as is in mobile.

Discourse works this way. It's a Rails-powered API with an ember.js front end.

https://github.com/discourse/discourse

I really liked where they were going with this and understood the idea of it rather fast.

I think its best seeing that you can also support and branch the API for that as well.

All your cache then sits on one api site rather than on multiple different sub domains...

Its an interesting concept.

I haven't found that there's much overhead at all. And there are a LOT of benefits... for instance, you can use a much leaner framework on the services side, since all you really "need" is routing. The services can (should?) run on separate boxes, so you can divvy up labor vertically. You can also version the API so that when you want to make changes, you don't have to upgrade your web app all at once.

FWIW, Trello didn’t launch with an API. It’s something we added in the first few months after launch. Just about everything goes through the API now since the mobile apps also use it, so you could say it’s an API first design NOW.

this blog post is by Bobby Grace, not by me.

My mistake. Feels similar in style to your articles :) Sorry Bobby.

It is one of the advantages of the API-first approach. I am convinced that's the way to go for building a modern Web app, but spinning off a special version of your client is not free; you still need to apply, as the author put it, "the occasional fix" and need to be aware that part of your audience is experiencing things through a different interface. Moreover, this approach requires a bit more investment upfront compared to regular Web development approaches.

The worse problem is IE8, which is the highest supported version on XP, and is a way bigger piece of crap than IE9... especially with XP's still high market share.

http://arstechnica.com/information-technology/2014/03/weeks-...

It's as if it was only yesterday that we saw Microsoft bragging about IE9 being arguably slightly ahead of the other browsers.

Ugh... yeah, I'd give my pinky finger to only have to worry about IE9 and up.

Windows XP is abandoned on April 8th. A large number of the Windows XP machines still in use are in China (nearly all pirated). The other ones can often be safely ignored for most businesses as they're not going to be customers anyway.

I'm not sure why you think that the non-Chinese XP machines won't be potential customers. In my experience, the largest remaining IE8 (and even IE6) holdouts are big enterprises and government organizations. Those organizations are also some of the richest potential customers out there.

For our product (enterprise task and workflow management), we see about 2% IE6/7 and about 5% IE8 in our customer base. Of those customers, most are planning to upgrade away from XP sometime in the next year or two.

That's why I said often. If you specifically deal with a US govt agency or large corporate customer with a poorly run IT department, you may very well still need those customers. But for most consumer businesses, Windows XP users aren't really going to be customers.

Internet time moves fast, but not that fast. No way is it that hard to support IE9, and no way that a three year old browser can be classified a dinosaur. IE9 has a modern fast ES5 javascript engine, sufficient css3 support, canvas, inline svg, and fixed most of the IE-isms in earlier versions. Unless trello is porting their codebase to webgl i cannot imagine it being that much work to support ie9. Supporting ie9 is simply part of the cost of doing business.

The article states support isn't being dropped, rather IE9 users are being shown to a land frozen in time. Not a bad strategy I'd think.

I don't understand why Microsoft haven't created a special application servear for old IE versions. If an Enterprise could install a server which provided IE via a Remote Application window it would allow them to upgrade to the latest version of Windows.

Remote Application is already used to provide support for legacy apps, much in the same way that terminal emulators are used to access legacy mainframe systems.

Microsoft provides free* virtual machines, each with a specific version of IE. Not quite the same thing though, but they are actually pretty cool.

http://www.modern.ie/en-us/virtualization-tools

* not sure if an enterprise can use them though.

If you go up to the main site, you'll notice there's a "free 3-month trial of BrowserStack". I think this effectively means that enterprise is fine.

IE9 is a dinosaur? Loads of locked-down corporate intranets still run IE6.

The fact that one is from the Triassic while the other is from the Cretaceous doesn't make the latter less of a dinosaur.

I can't think of any dinosaurs that first came into existence 3 years ago in 2011 and were the most recent dinosaur of its particular kind available until September 2012, 20 months ago. It seems like a terrible metaphor to say everything more than 20 months old is a "dinosaur". It's such a bad metaphor in fact that I wonder for the sanity, competence and/or neutrality of anyone seriously advocating it. I doubt anyone genuinely and sincerely believes that things more than 20 months old are all dinosaurs, even tech things. On the other hand, I would not be surprised at all if such terms were used as part of a propaganda or agenda driven attempt to mislead people. That sort of thing happens a lot.

This seems rather hyperbolic. What sort of agenda do you imagine I have that would make me want to propagandize and mislead people? Am I part of a secret cabal that seeks to enslave the sheeples by encouraging them to keep their web browsers reasonably up-to-date?

(For clarity, I wouldn't normally describe a two-year-old browser as "a dinosaur," mainly because the 5-year-old inside me thinks dinosaurs are too cool to use their species as a pejorative. But I would describe one like IE 9 as "fairly outdated," which is just a less goofy way of saying the same thing.)

Well, it may not be a dinosaur in terms of age, but it certainly is in terms of its featureset. Given IE's track record, the fact that it was the latest version 20 months ago doesn't mean it was a GOOD version 20 months ago.

> IE9 is a dinosaur? Loads of locked-down corporate intranets still run IE6.

IE6 is pond scum.

Interesting to see that users using Vista are already being seen as 'should be dropped'. AFAIK Internet Explorer 9 is the highest IE version on MS Vista.

Not that I like IE or Vista, just seems to me that there are a ton of people out there with IE9 as max IE version.

It makes me wonder why Ms bothered tying the browser to the OS in this way. It's not as though the Justice Dept. is going to haul them back into court to explain their fibbing about how great it was to tightly couple the OS to a particular "strategic" executable that runs on it. Was that really a concern when they came out with Vista?

How do they determine what browser version users are on so they can show the right version of jellydoughnut-xx?

He kind of skips over this, but the way he says you don't even get served the "collection of files" until your browser is determined makes me think they're using the request header.

So maybe that 8.48% decrease is just people starting to spoof.

I feel like if you're savvy enough to be using Trello and yet still on IE9, there might be a reason you haven't upgraded.

What, so you spoof your client so you can get a version that's unusable in your browser because of its poor feature support? That doesn't sound like a great idea.

If you're desperate for new features, you might not care if the CSS is a little wonky.

That's not a good idea, for the record. We dropped IE9 so that core features could drop IE9 support, not to save a little CSS work. In the near future Trello will become quite unstable in IE9.

How many people who use trello even knows how to get IE9 to spoof a UA header? I sure don't. And if they can't update their browser, they probably can't put some sort of request proxy in between to change headers.

>still on IE9, there might be a reason you haven't upgraded.

Corporate IT users sometimes dont have a choice which is the case for some users on my current project. 2011 isnt that long ago to support something for a corporate app but I guess things move much faster other places.

That's my point -- if you haven't upgraded by now, you might not be able to. Period. Which is not to say I think they should fully support IE9. I'm just saying that (maybe misguidedly) I expect Trello users to be a little smarter than the average bear.

Reply on news.ycombinator.com