I suppose you must be referring to character entities.
If you use the most appropriate encoding standard for the language with which you are working, you should only need to use entities for symbols (and not necessarily all of them at that). The most useful (and the most widely supported) encodings are the ISO 8859 series and UTF-8. If you use UTF-8 (which I and many other coders do) it is good practice to avoid including the BOM (byte order mark) while it is still not universally recognised. As legacy browsers and platforms die out, such issues go away. As always, test thoroughly across as many common platforms as possible: any problematic characters will become obvious.