It is just 17000 records of just one field CustomerID.
I am using everything that people mentioned above for different projects: Linked server with the text driver, OPENROWSET, CLR in 2005 to deal with text files, BCP, an independent (from the third-party) Reporting Database. I also support MS ACCESS /EXCELL VBA import from text files, VBscript from the ASP pages, VBscript by the cscript, whatever. My preferences are to use the back-end processing on the same server after getting the source data into temp tables. It is usually faster then doing the front-end processing or distributed queries.
It would have been good to mention that in SQL 2005, you can achieve the same result much more easilly by using a simple Merge Join Data flow transformation.
How about using SQL2005 express on my PC (local installation), use DTS to import the customer ID into a new table then T-SQL join the customer ID table to the linked db in SSM (provided that the db is linkable)?
Is there any problem or disadvantage doing it in this way?