The only check for whether a user is allowed to change the compression state on a file is whether they have read or write permission. If they have write permission, they can change the compression state locally or across the network.
The two ways to measure the performance of NTFS data compression are size and speed.
You can tell how well compression works by comparing the uncompressed and compressed file and folder sizes. The earlier section of this chapter titled "Using My Computer or Windows NT Explorer" describes using the Properties tab to view the compressed size and compression ratio of a selected file. The section "File System Utilities" in Chapter 22, "Disk, File System, and Backup Utilities," describes using the DirUse program to see the compressed size of folders.
Using NTFS compression might cause performance degradation. One of the reasons this might happen is that, even when copied inside the same computer, a compressed NTFS file is uncompressed, copied, and then recompressed as a new file. Similarly, on network transfers, the file is uncompressed, which affects bandwidth as well as speed.
With data compression, the question is "How will it affect performance on a computer running Windows NT Workstation or Windows NT Server?"
The current implementation of NTFS compression is definitely oriented toward Windows NT Workstation. Compression on a computer running Windows NT Workstation does not seem to produce a substantial performance degradation. No one who has started using NTFS data compression at Microsoft has lodged any complaints about performance degradation.
A computer running Windows NT Server can be another story. Some normal production servers at Microsoft (source servers and binary release servers) have been converted to use NTFS data compression, without any complaints about performance. However, using a tough server benchmark, like NetBench, shows a performance degradation in excess of 50%. Read-only or read-mostly servers, or any lightly loaded servers, may not see a severe performance penalty. Heavily loaded servers with lots of write traffic are poor candidates for data compression, as shown by NetBench.
You really need to measure the effects of data compression in your own environment.