When a dataset is deposited in a repository, it is given an identifier: a code which is unique to (this version of) the dataset. This code makes it possible to cite this specific dataset, for example in (your own) publications.
The identifier of a dataset may be compared to the ISBN of a book, with this difference: the ISBN is not included in citations of a book, whereas the identifier is a fixed part of a reference to a dataset.
Different kinds of identifiers are current. Some of these identifiers (DOI, Handle) are used not only for datasets but also for other electronic publications.
The Uniform Resource Locator (URL) is the address of a web page, for example a page on which a dataset is made available.
The use of an URL as identifier is not without risk: the web address that you see in your browser often contains data which are valid only at that moment (session bound). If you save this web address unchanged and later open the page again, you are likely to get an error message.
If you have a choice between a URL and some other identifier, for example a DOI, it is better to use the other identifier so as to have a trustworthy link.