Plenty of broken image links

Question

This bug was detected in 2011: the original policy was to use software centre images from bit. ly/ software-small, but that mysteriously broke for some reason that doesn't really matter at this point. A quick query points to >500 posts that still link to this image, and which look sort of like this:

The answer to that bug report suggests that the image files to use should be http://hostmar.co/software-small and http://hostmar.co/software-large, which will "always" link to the appropriate Software Centre icon. This is quite obviously false, since e.g. in this other meta post you can see such delights as

More generally, hostmar.co is effectively a url shortener and that puts a huge fraction of posts at the mercy of service interruptions on Marco Ceppi's provider. Marco Ceppi stated back in 2011 (yeah, that's five years ago) that

Stack Exchange is aware of this and a fix is in the works with no ETA at this time.

I understand the value of having a self-updating icon, but an external hosting site is not the answer. Changing from one url shortener to another, less-well-known one is, frankly, just building fragile infrastructure. (In particular, note that the hostmar.co links are not robust: on my machine, the banshee link above oscillates from showing to broken. Or, put another way, the fact that you've changed a link to hostmar.co and that it shows on your machine is not a guarantee that it will show for other visitors, even at the same time.)

Given that this is an icon that only needs to be updated once every many months, for me there are two obvious solutions:

• change all the bit.ly and hostmar.co links to i.stack.imgur versions via a dev-side query, and then update as appropriate, or
• ask SE to add the image file to the site's graphic content (possibly as part of the sprites sheet, though that does require a bit more development)

Either way, this site has over 500 posts with fundamentally broken layout. Is this something that this site considers to be a problem? My inference, given that it's been obviously broken for five years, and that this can be quickly fixed dev-side by a quick query without dumping hundreds of posts on the front page, is that this sort of mangling isn't considered worth fixing. If that's the case, this site is pretty unique in the SE network in terms of how much it cares about its content. On the other hand, maybe I did miss out some systematic campaign to fix those hundreds of broken posts, which isn't documented on this meta.

Just sayin'.

Answer 1

The problem with your first proposal is that it becomes impossible to update the image the i.stack.imgur.com URL points to. It is also worth pointing out that imgur.com URLs are not easy to memorize.

I would be in favor of your second proposal, however. Or, even better, perhaps they could add a Markdown tag that automatically expands to the correct URL. Something like this, maybe:

[package:roundcube-plugins]

We already have a similar syntax for tags.

Answer 2

While there's a lot of improvement that can be made here, any changes will be slow to adopt. I think your ideas are great and echo a lot of what has been discussed in the past. The occasional image loading failures were due to an ageing server which is running the hostmar.co domain. It's been migrated, and pending DNS, will be resolved in the next 48 hours.

Furthermore, I'll make sure the rest of the moderation team has access to this machine to help manage logs and any errors that might occur going forward.

From what I recall several years ago it was more than just a simple find replace, the Stack Exchange team detailed what seemed to be quite a bit of work to replace every entry of X and was only reserved for drastic measures. That's where the hostmar.co alternative came from. It happened to be a domain I had at the time.

We can, in addition to your suggestions, also consider just building out what runs hostmar.co (which is now just a set of NGiNX rules - was a php script). Giving us the same capabilities but with a more crisp domain name.

Answer 3

Shog9 has nicely run his link updating tool for us, replacing some 60 or so links. For some reason there are still a few instances that were not indexed by SE search and so the script could not fix those. This may have been due to the ongoing DDoS attack at the time. I will check again in a few days and see if it's been fixed. Otherwise we'll have to manually find and fix the rest.