In my Amiri font I used to have dots separated from base of the glyph, the idea is simple:

* first you have a ccmp feature that decomposes the glyphs, something like (using feature file syntax):

feature ccmp {
sub haa by haa.base;
sub khaa by haa.base dot.above;
sub jeem by haa.base dot.below;
...
} ccmp;

Where haa.base is the dotless glyph, and dot.above and dot.below are zerowidth glyphs with mark glyph class.

* from now on all glyph substitution is done on base glyphs, e.g.:

feature init {
lookupflag IgnoreMarks;
sub haa.base by haa.base.init;
...
} init;

feature liga {
lookupflag IgnoreMarks;
sub haa.base.init haa.base.fina by haahaa.isol.liga;
...
} liga;

The IgnoreMarks flag is very important, else the dots will interfere with the substitution.

* lastly, you use anchor marks for placing dots just like tashkil marks.

* some characters have different dot placement depending on their form, e.g. farsi yeh have dots only in final/isolated forms, for these characters you are better doing decomposition in init, medi etc. features, e.g.:

feature init {
sub farsiyeh by yeh.base.init twodots.below;
...
} init;

feature fina {
sub farsiyeh by yeh.base.fina;
...
} fina;

However, this has some side effects, e.g. sometimes you need different kerning for glyphs with dots below than glyphs with dots above, sometimes certain dotted combinations don't go well together and you need to avoid ligatures in these combinations and this becomes tricky because the dots are treated as combining marks so you can't use them as context until you remove the IgnoreMarks flag and I ended up abandoning the idea.

* A different strategy to decrease the number of glyphs is to avoid ligatures at all and use contextual forms instead. e.g. instead of the haahaa ligature above you "cut" the ligature into two parts:

@clsHaaInit = [haa.init khaa.init jeem.init ...];
@clsHaaFina = [haa.fina khaa.fina jeem.fina ...];
@clsHaaInit.HaaHaa = [haa.init.haahaa khaa.init.haahaa jeem.init.haahaa ...];
@clsHaaFina.HaaHaa = [haa.fina.haahaa khaa.fina.haahaa jeem.fina.haahaa ...];

lookup haahaa.isol {
sub @clsHaaInit by @clsHaaInit.HaaHaa;
sub @clsHaaFina by @clsHaaFina.HaaHaa;
} hhaahaa.isol;

feature calt {
sub @clsHaaInit' lookup haahaa.isol @clsHaaFina' lookup haahaa.isol;
} calt;

Since we have 15 haa-like characters in Unicode, this cuts the number of glyph needed for this ligature from 15×15=225 (number of all possible combinations) to 15+15=30 (number of components).

(PS. the names used above are arbitrary, you can use whatever glyph names you like).

Yes, this is possible, as Khaled describes. One thing to consider, though, is that this will break with Microsoft Office's options to display marks in different colours or not to display marks, since it has no way to distinguish dots treated in this way from other marks. This shouldn't necessarily stop you from doing this approach -- and for some styles of Arabic such as nastaliq it is the only sensible approach --, but it is something to be aware of and may result in some support contact from confused users.

Thanks.

It solved the problem. I am grateful to Khaled Hosny for such detail and taking time to explain the solution with an image. I also appreciate John Hudson for providing valuable piece of information.

Once again Thanks.

Thanks Khaled Hosny for the detailed explanation.