There isn't a simple answer to this question, because -- yes, the OSR tends to be a set of values / principles. If you are looking for a strict taxonomy you aren't going to find it.
Here are some definitions that I find useful, and reflects the stuff I like about OSR-style games / design / play:
" If you want a 'ten commandments' then this breakdown from the famous Scrap thread works as well as any;
1. This is a game about interacting with this world as if it were a place that exists.
2. Killing things is not the goal.
3. There is nothing that is "supposed" to happen.
4. Unknowability and consequence make everything interesting.
5. You play as your character, not as the screenwriter writing your character.
6. It's your job to make your character interesting and to make the game interesting for you.
7. If you find yourself in a fair fight, your tactics suck.
8. The answer is not on your character sheet.
9 .Things are swingy.
10. You will die
By Gregory Blair, Brian Harbron, FM Geist, Zedeck Siew, Brian Murphy, Dirk Detweiler Leichty and Daniel Davis;
There's not deep theory here, I'm just trying to describe a 'scene' or social/creative web that I intuit more than see around me. "
(from this post by Patrick Stuart; the list was crowdsourced from a discussion thread on G+)
~
Recently I saw Sean McCoy (who made Mothership) cite Ben Milton (who made Knave)'s definition of the Old School:
" OSR: The more of the following a campaign has, the more Old School it is: high lethality, an open world, a lack of pre-written plot, an emphasis on creative problem solving, an exploration-centered reward system (usually XP for treasure), a disregard for "encounter balance", and the use of random tables to generate world elements that surprise both players and referees. Also, a strong do-it-yourself attitude and a willingness to share your work and use the creativity of others in your game. "
~
Speaking personally, I like games that don't try to model "rules of narrative" stuff:
Ie: Mothership is a sci-fi horror game, and this is emphasised by how gear in the game works, how squishy the PCs can be, how easily it is for PCs to freak out. But, depending on dice, it is still possible for a game of Mothership to veer "off-genre" -- it could be a funny comedy of errors, or it could turn out to be a Doom-like shoot-em-up.
That's by design, I feel? in my experience, games built on OSR principles tend to be foundations -- "Here are some tools to help you imagine being in a thin vaccsuit with maybe a terrifying alien stalking you; go forth and have an unpredictable few hours in a shared imagined space with friends." -- as opposed to scaffolds -- "This game is about space horror, and your group will the narrative tools to have an experience similar to other kinds of space horror media".
This isn't a mutually-exclusive dichotomy, more a matter of emphasis.
(I think I first saw games described as "foundation" or "scaffold" by Mabel Harper; I'm not sure whether I'm using this framework as she intended, or misquoting her, though.)
~
Also OSR-y folks like Emmy Allen and Luka Rejec have touched on the idea that D&D / D&D-esques are a shared language. This feels true, to me. The Old School tends to use the familiar grammar of D&D (and other well-known rulesets) to write new and fresh things, and converse with each other. This is why Daniel Sell's initiative system for Troika! (as one example among many) is something you can easily jury-rig into other things.