TL;DR: readers should not poll more often than once and hour, use ETag and If-Modified-Since to determine whether to download the full feed again.
Which items you have seen previously is something the feed reader keeps track of.
TL;DR: readers should not poll more often than once and hour, use ETag and If-Modified-Since to determine whether to download the full feed again.
Which items you have seen previously is something the feed reader keeps track of.
It really depends on you but IMO for most feeds polling once a day is plenty.
Is there a particular field that can be used as an identifier?
Of course a lot of feeds also get this wrong and change the GUIDs for existing entries once in a while which results in strictly compliant readers showing you the entire feed history as new. Really annoying.
The latter is annoying, I agree.
¹ It is an NNTP interface so the article is superseded; https://feedbase.org/about/ - if you don't want to see updates, you can configure your newsreader to skip supersedes.