This is not a silly issue at all, but extremely important.
The intent of the PS3.10 file meta information is to describe how
the data set is encoded, and who wrote it (for traceability).
To that end, best practice is to DISCARD any file meta information
that may have been read in, and write only what new file meta
information is required, known or relevant to the writer.
The writer should know, and should populate de novo:
File Meta Information Version (always 00H, 01H; get the byte order right)
Media Storage SOP Class UID
Media Storage SOP Instance UID
Transfer Syntax UID
Be especially carefully, obviously, if you have "changed" the
SOP Class or Instance UIDs or encoded in a different Transfer
Syntax than the original file was read in.
The write should also have their own:
Implementation Class UID (should not be copied from someone else)
And should RECALCULATE:
File Meta Information Group Length (since it may be different than read)
since the length will have almost certainly changed.
Everything else is optional, and the AE Titles have a specific
meaning, so read carefully:
http://dicom.nema.org/medical/dicom/current/output/chtml/part10/chapter_7.html#sect_7.1
Specifically:
Source Application Entity Title (is you, the writer, if you have one)
I would strongly recommend not "copying" (and preferably not populating
for your own nefarious purposes):
Private Information Creator UID
Private Information
either, since you have no idea whether what was in there was specific
to the original file "layout" or "content" or not (e.g., it might
contain fixed byte offset dependent information), or be dependent
on content in the data set that may have been changed). It is rarely
present anyway, so no harm will be done by discarding it as a routine.
David