I use only classes and never use IDs. Many people like to use IDs for different reasons.
I've seen many questions regarding IDs vs classes on stackoverflow, but no one addressed pure code organization point of view disregarding compatibility and runtime performance.
From code organization point of view, I think that using IDs is bad just like using global variables in Visual Basic code.
One reason is that IDs have to be unique which introduces unnecessary and bad dependency between different independent parts of your code (controlling different parts of HTML DOM tree).
Another reason is that making new class names is actually easier than ID names because with IDs you have to worry about global scope and with class names you need to worry only about uniqueness in local scope, same benefit as with local variables.
Most people will argue that performance of addressing by ID is better than by class and I will agree with that. But as browsers become more advanced with native implementations of CSS addressing from javascript and computers become faster, performance becomes less and less important. So let's disregard it and concentrate only on organization of code in context of current question.
This discussion started here, but my potentially wrong advice generates negative points and became too big to keep in comments, so here I try to convert it into something positive and manageable.
One visible point in favor of IDs is to use them as a tool of rule prioritization because priority of #name is higher than priority of .name.
My response: using IDs to raise priorities is bad hack, it's cleaner and there is more freedom if you use additional root elements inserted between body and other levels of tree, for example priority of body div div span.class1{} is higher than body div span.class1{} is higher than body span.class1{} is higher than span.class1{}. Another tool to use for that purpose is !important.
Some may argue that using more root elements means more difficulties when the page structure changes, but I don't think this is the case because you never have to put anything between body and designated for prioritization divs. Those divs can always stay below body and above all other content.
Another interesting association was brought about pointers and that IDs are not bad because pointers are not bad.
My response: pointers are bad if you hardcode absolute memory address in your code. Using relative pointers is always better (examples: using segments(CS,DS,SS,ES) in 8086 CPU; relative variable and method addresses generated by compilers). If we consider DOM tree as memory and compare using ID to using class then #name represents absolute memory address, but div.tab1 .name represents relative address (relative to div.tab1).
Another supporting point that I've seen for IDs is that elements with IDs are more easily available in javascript as becoming global properties. My response: again, this is like saying that global variables in Visual Basic are more conveniently available. The problem is that you can't keep large enough global (or any other) namespace in order without introducing naming hierarchy like level1_level2_name, which is just a hack to replace one namespace mechanism with another. DOM tree is convenient enough to organize namespaces, why disregard it ?
Namespace simulation inside IDs using underscore is bad because you can't establish naming context and will have to duplicate all paths everywhere in your code. That practically means that you won't be able to use CSS preprocessors that fix inability of CSS to use contexts.