Tibetan has been encoded as a character set since the mid nineteen nineties, and the decisions taken then shape how the script behaves in every piece of software that handles it.
Two of those decisions matter to a reader. The first is that stacking is not encoded as pre-composed units. A syllable with a superscript, a root, a subscript and a vowel is stored as a sequence of separate characters, and the visual stack is assembled at render time by the font. That keeps the character set small and searchable, and it puts a great deal of responsibility on the font: badly rendered Tibetan is common, and it is usually a font or shaping problem rather than a data problem.
The second is that dependent marks, the four vowel signs among them, are characters that have no standalone form. To display one in a chart, an isolated mark is shown on a dotted circle, which is a rendering convention and not a Tibetan letter. Anyone who has copied a vowel sign out of a code chart and found a small dotted ring attached to it has met this.
The encoding also keeps distinctions that a casual reader might expect to be merged, including a separate fixed-form ra for the cases where the ordinary letter must be prevented from stacking, and separate characters for the reversed letters used in Sanskrit transcription. Both are the right call, because both pairs are contrastive.