Re^8: An Idiot's Guide to YAML

[Mutant]

These days I view YAML as an opaque serialising mechanism. You look at it, but don't touch it.

I agree, it's not easy to make large scale edits, but it is useful if you just need to tweak the odd value here and there. I think the readability of XML is a real issue, which is one reason we've seen the rise of JSON.

[mr_mischief]

The issue of the readability of XML is real, but it's at least a one-time investment to get good at reading it. How many YAML-like solutions are out there in different circles, with library support in one language and not another? JSON, specifically, is nice because it's another standardized (ECMAScript) language with broad support. IME, XML is pretty easy to make one-time edits against in a text editor, just as people claim for YAML.

Now, in a large organization where a standard configuration language among many programs needs to be more lay-person readable than XML, there's definitely something to be said for teaching a bunch of people YAML or something similar. I'm still inclined to say that programmers or data specialists are more likely to want to leverage what they know in XML or in the implementation language than to learn a new, less standard data interchange format.

Not specifically to bring yet another format into the discussion, but I don't see any benefits of YAML I wouldn't get with LDIF, which can be directly imported into most LDAP implementations. If you're wanting to share a common config format among an organization, LDAP and LDIF seem ideally suited. Since that's the best scenario I can think of for YAML, I'm again at a loss to see the need.

Having options is nice, so it's probably good that there is a YAML for it to be another option. It's easy to have too many options to be able to evaluate them all fully, though, and to have too many different options chosen within different parts of an organization. If one is going to choose a standard, it's often smart to choose one which can leverage existing knowledge and existing tools. XML, LDAP (with LDIF), JSON, etc. seem to fit this bill better than YAML.

Now, there's two separate arguments here. One is whether YAML is good enough and offers important enough advantages to be a widely accepted and supported standard. The other is whether or not it already is a widely accepted and supported standard. If the former argument passes muster, then some supporters of YAML need to work on satisfying the latter. Until both are satisfied, either argument by itself is going to be answered with 'Yeah, but...' and "So what?' types of comments.