>My client does not want to 'print directly to the printer' because they had
>a few bad experiences with the printer going offline while the negine is
>running.
>
>Is there a way to determine whether the local spooler has pushed the data
>to the network spooler? We could delay the printing of the current document
>until all data has been pushed to the network spooler.

There must be - if you double-click on an installed printer you get a (semi-) real time display of the printer/spooler status. However, I have no idea what API(s) are used to obtain this information.

>
>We will try to output print different printer
>
>Printing to a file. Is there a practical limit to the size of the print
>file to be efficient? The engine has two usages. (1) it processes a data
>file that is to be printed after hours. This data file typically contains
>thousands of records (up to few hundred thousands on a couple of weekends a
>year) and the documents are printed on an IBM InfoPrint 40 printer. We are
>not aware of any problems with this after hours usage. (2) The engine is
>also used to print letters on demand. The user selects up to 100 documents
>and starts the engine. The documents are printed on a HP LaserJet 5. This
>is the usage that gives us some problems.

I'd suggest print one document to one file, named with something like SYS(2015). That way none should get too large. As for efficiency, as long as there's some write caching in the file system and the file size(s) don't exceed the disk cache size you may not even go out to disk before you're finished with a given file, so efficiency could be pretty good.