Improving Interoperability by Short Release Cycle

When a software is shipped, it has bugs. There are many reasons for these bugs. It can be poor in-house development, it can be careless testing, it can be unclear specifications, and many other things. We have to live with these bugs in software.

A bug deployed in a software for a long term becomes a feature.

It's specifically true in a distributed environment where pieces are loosely joined: the Web. Softwares are released with their inherent bugs. Content and framework developers are hit by the bug. They modify their own software to accommodate the bug or take advantage of it. No new version of the buggy software is released for a long time. When it is finally time to release a new version, the buggy software has to keep the bug as a feature to not break anything on the Web. Eventually, one day the bug makes its way to a specification like html 5.

It is difficult to change things because they are all intertwined but in a very loose way, which makes its strength. You can try to fix the software knowing that it will break things at many places. You have then to be ready to loose customers if someone else as implemented the bug. Users are not aware of the bug, and they don't really care about it. Fixing means also, in this case, educating people about the issue, and content developers on how to fix their content. Content developers will be the hardest ones. If they fix, knowing that it will break things in other softwares, they will loose customers. So they are not likely to do it.

To avoid that bugs become features, softwares have to be released with a short cycle. So that people can't take advantage of bugs. It means also that bugs don't survive many releases.

Can we improve the situation for bugs already deployed?

The solution could be a simultaneous release of softwares and a campaign educating people. This is challenging. Very challenging. It means agreement between companies at the release moment and a front with regards to unsatisfied customers. I just wonder if it would be possible as an experiment for one or two bugs. For example, in HTML 5 specification, browsers and Web sites, would it be possible to fix the content-type sniffing on text/plain.