@Kasper Just noticed your question on the main site. Following our discussion I've been watching myself using unicode in writing an article. I'm getting to the point where I'm considering the experiment over and that it isn't worth the hassle. The gain in "readable source code" is - for me - a myth. LaTeX is already readable enough and the unreadable parts are not made more readable by using unicode symbols. The problem is that not everything is covered by unicode ...

... and so there is a mish-mash of unicode and macros which actually makes it less readable as there isn't a consistent style. Here are some samples to show what I mean:

  ∑_{k=0}ⁿ a_k (ψ - 1)^k = ∑_{k=0}ⁿ ∑_{j=0}^k a_k \binom{k}{j}
(-1)^{k-j} ψʲ

So here we have a mixture of super and subscripts. Some can be represented by unicode symbols, some can't. Now, it may be that ^k can be and I just didn't get the right input for it, but ^{k-j} can't, nor can _{k=0}. Moreover, the ⁿ seems to get a little lost against the crowd.

There's also that \binom{k}{j} in there which now jars because it is the only macro present.

 \{1,ψ, \dotsc, ψⁿ\}

In this one, again there's a macro \dotsc which looks out of place. Also, that ψ is a little tricky to differentiate from φ.

So it's like I've done a partial compilation and some things have been converted to their final form but not everything. And that makes it harder to read (for me).

Contrast this with unicode accents: Fr\'echet versus Fréchet. Now the unicode really does make it easier to read.

@AndrewStacey I understand what you mean. But I don't think that unicode will ever cover everything, and I don't think it should. I think it would be better to have an editor that supports more opentype layout features.

@AndrewStacey Hm.. I don't think this is completely right. I understand what you mean with those superscripts and subscripts. I just always use _ and ^ to remain a consistent style.

@AndrewStacey Isn't the same true for $x≤2⇒y≥3$ is more readable then $x\ le 2\to y \ge 3$ no ?

@Kasper Well, no. It's hard for me to read the symbol you've used for \to (I suspect it is actually \implies). Indeed, I would say that x \le 2 \implies y \ge 3 is more readable in part due to the fact that \implies is spelled out. So I'm writing my mathematics by concentrating on what the mathematics means instead of how it should appear.

@AndrewStacey Dyslexic persons like me won't ever agree with that, but I understand what you mean.

@AndrewStacey But if macros are really more readable, why not use \equal instead of = ? I think that if you prefer = above \equal, than it would be logical to prefer ≤ above \le.

@AndrewStacey You could argue that it's not consistent to write math symbols like =,+,-,<,> as an symbol, but all other math symbols as macro.

@Kasper On the other hand, dyslexic persons like me will say that the distinction between \to and \implies is easier to see because it is greater than between a single arrow and a doubled arrow.

@AndrewStacey You are dyslexic as well ? Well, I thought dyslexic persons prefer symbols/numbers way above words in general, but this doesn't seem true.

@AndrewStacey For me, I understand the meaning of a symbol in a milisecond, and the meaning of a word takes much longer.

@Kasper Consistency is not a reason for doing anything. Consistency is a "when everything else has been taken into account, I'll be consistent". Certain symbols are easy to type because they are readily accessible on any keyboard and for those, I'll use the symbols. But my point is that due to the vast enormity of symbols one has to draw a line between using symbols and using names. Exactly where that line is drawn will largely be a matter of taste and experience.

@Kasper I'm a mathematician who's been writing LaTeX for over a decade. Names are no problem. And my dyslexia is weird so I wouldn't take any notice of that, I put it in essentially to say that being dyslexic is orthogonal to preferring symbols over names.

@AndrewStacey But you said that you draw the line at not using unicode at all, because there isn't an consistent style.. But I don't understand this, because you don't have an consistent style now either. I don't understand that you (and every other mathematician I spoke about this) don't seem to see any value in those unicode symbols. Ok, supersscript, subscripts, that I understand, but I think that symbols like ≤,≥, would be improvement anyway in the long end, and make the code more consistent.

I mean writing this is $x<2 \implies y \le 3$ looks to me as inconsistent as using for some letters supscripts and for some letters ^

@Kasper I said that for me, I would draw the line in mathematics at "what I can easily type with my keyboard" so if it's ascii, I'll use it, if not, I won't. That doesn't mean I don't see the value in unicode! The point is that my source code is mine (and my collaborators', which is not a negligible consideration) and so I can write it how I like. I've gotten used to a certain way of writing and changing it requires evidence of how it makes life better for me (ctd)

(ctd) my source code is not meant for public view (not that I hide it, I just don't care about what others think of it). It's what it looks like after compilation that I really care about.

@AndrewStacey I'm not trying to convince you to use more unicode, I'm just trying to find out why you prefer not to use it.

@AndrewStacey For example, I don't understand this. Your editor was able to automatically converts \le to ≤ right ? I don't understand why you change that back to not converting \le to ≤ ..

@Kasper I understand that.

@Kasper Because either I had to set it to convert \le (note the space) or had to keep correcting it when it converted \leftarrow to ≤ftarrow. So it's easier for me to not bother with converting to unicode. It is, though, nice to know that if I had something that already used unicode then I wouldn't have to convert it back.

In effect, I'm saying that unicode is (for me) more of an output format than an input format. My keyboard constrains me to type in ascii (more or less, it's a norwegian keyboard so I get åøæ as well). Any system for using unicode means that I have to have something that converts the ascii that I type on my keyboard into unicode. TeX already does this just fine, so I have no reason to introduce another step into the chain when I don't get any benefit from it.

@AndrewStacey Ok, that I understand.

I should say that I also have more inventive uses for the unicode-math package - do a search through my questions on the site to see what I'm on about. So while I might not type unicode, I appreciate its existence!

Well, good to hear I'm not the only one :)

@AndrewStacey I was chatting in the math room yesterday, and I've got the feeling that I'm the only mathematician in this whole world who prefers unicode math symbols above latex macros. I understand your reasons now. But I still don't get why not only most mathematician prefer latex macros, but most even seem to be heavily against the use of unicode in latex.

@Kasper Probably because most mathematicians have at some point emailed their LaTeX source code to a collaborator who knew even less about encodings and similar than they did.