The problem is the Base64-encoding which, when done in one go with moderately large amounts of data, needs temporarily a lot of space due to the creation of integers with ⎕UCS as an intermediary step.
If a WS FULL occurs the encoding should fall back to a loop.
The Laguntza package is an example: it causes a WS FULL with under 1MB of data (!) with a WSSIZE of 64MB of which 45MB are available (!) when the encoding is initiated.
The problem is the Base64-encoding which, when done in one go with moderately large amounts of data, needs temporarily a lot of space due to the creation of integers with
⎕UCS
as an intermediary step.If a WS FULL occurs the encoding should fall back to a loop.
The
Laguntza
package is an example: it causes a WS FULL with under 1MB of data (!) with a WSSIZE of 64MB of which 45MB are available (!) when the encoding is initiated.