https://bz.apache.org/SpamAssassin/show_bug.cgi?id=7358
Mark <[email protected]> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |[email protected] --- Comment #10 from Mark <[email protected]> --- As a rule of thumb, scanners should support what popular user agents support and user agents should support RFC's. RFC's are not leading in this case. Even though a boundary should be unique, we know that it sometimes isn't and we know that popular user agents support it anyway. Therefore scanners should support it or else risk detection failure. This is the multilayered MIME structure of the sample message: multipart/mixed (boundary1) | |-multipart/alternative (boundary2) | | | |-text/plain | |-text/html | |-boundary2 close | |-application/ |-boundary1 close A common way to parse this is to follow the layered MIME structure of the message. When multipart/alternative parts have been handled as a self-contained section, continue using the original boundary to find more MIME parts. Even if the boundary in both layers is identical, a scanner parser should know parsing isn't done after parsing the /alternative MIME. Closing boundaries can be missing sometimes, especially the last one in the file. MIME can be nested deep. -- You are receiving this mail because: You are the assignee for the bug.
