I still don't get descriptive notation

I don't get descriptive notation 

Descriptive notation does not use coordinate geometry and so is actually easier to keep the moves in the head. 

It's easiest to understand in talking about endgames, where it can be superior to algebraic.  In algebraic, if I want to talk about an advanced pawn on the edge of the board, I have to say "A White a- or h-pan on the sixth rank, or a Black a- or h-pawn on the third rank."  In descriptive I can say "A RP on the sixth rank" and it covers all of those possibilities, because the rank names are relative to the moving player.

People actually revert to talking this way quite often--"doubled rooks on the 7th rank" when it should be "the 2nd or 7th rank"  and "a RP" rather than "an a- or h-pawn."

So a quick rundown:  The files are named after the piece that originally stood there.  So QR file, QN file, QB file, Q file, K file, KB file, KN file.  The ranks are named starting from the side of the moving player, so White's pawns are on White's second rank (but those same pawns are on Black's seventh rank).

Then you cut off as much extra information as you can, to make the move notation shorter.  So if you have two knights on their starting squares and you move one to f3, you say N (or Kt) - KB3.  But if it's unambiguous (say, the other knight is gone) you say N-B3.

Captures focus on what you are taking, not where you are taking it.  If only one pawn can take a knight, PxN.  People often talk like this too--my kid students often say "QxR+" when they mean "Qxd1+" because I think it's a very natural thing to say. (Or they caught it from me.)  If two different pawns can take a knight, you have to say which:  QPxN, NPxN for a knight on c3 which could be taken by either the b- or d-pawns.

One awkward thing is remembering which knight is the QN and which is the KN after they have hopped around a bit.  You can instead disambiguate N(KB3)-KN5 when the other knight could have gone there too.

Does this help at all?

It's wordier than algebraic but I do kind of miss it for endgame analysis.

Just like any language, descriptive holds a meaning and can become easy to use when you practice it, but can be rather confusing when you don't practice it.  And just like any language, it has its awkward elements too.

Mary made a very good case for it -- better than I've ever seen or thought about.  Thank you, Mary!

I would say that algebraic notation is easier to keep in mind once you have mastered coordinates (but that's easy for me to say since as a mathematician by training I never had problems with coordinates ).  At the same time, the descriptive notation is "modal" -- i.e., what you say depends on the situation (the mode): you may write or omit some details -- and this characteristic precisely makes it harder to pick up, in comparison.  People who study intuitive user interfaces tend to agree that the presence of modal elements can be confusing, so they try to eliminate modality and make everything have one and only meaning.  For an illustration of the opposite, imagine if you had a button in the car steering controls that works as a break in some situations and also as an air conditioner in others ;-) -- obviously an unrealistic example, but it goes to show that with this modal interface the mind of the car driver will have to always decide which is the correct way, or else...

Another interesting detail is that, to my knowledge, descriptive notation only ever existed in the West (US only?).  Even old chess books in Eastern Europe, where I come from, had algebraic (or figurative) notation, but not descriptive.  This makes descriptive notation additionally confusing for people who have first picked up "the other way" (and the currently dominant way of algebraic notation).