What is the difference between /opt and /usr/local?

Question

According to the Filesystem Hierarchy Standard, /opt is for "the installation of add-on application software packages". /usr/local is "for use by the system administrator when installing software locally". These use cases seem pretty similar. Software not included with distributions usually is configured by default to install in either /usr/local or /opt with no particular rhyme or reason as to which they chose.

Is there some difference I'm missing, or do both do the same thing, but exist for historical reasons?

Answer 1

While both are designed to contain files not belonging to the operating system, /opt and /usr/local are not intended to contain the same set of files.

/usr/local is a place to install files built by the administrator, typically by using the make command (e.g., ./configure; make; make install). The idea is to avoid clashes with files that are part of the operating system, which would either be overwritten or overwrite the local ones otherwise (e.g., /usr/bin/foo is part of the OS while /usr/local/bin/foo is a local alternative).

All files under /usr are shareable between OS instances, although this is rarely done with Linux. This is a part where the FHS is slightly self-contradictory, as /usr is defined to be read-only, but /usr/local/bin needs to be read-write for local installation of software to succeed. The SVR4 file system standard, which was the FHS' main source of inspiration, is recommending to avoid /usr/local and use /opt/local instead to overcome this issue.

/usr/local is a legacy from the original BSD. At that time, the source code of /usr/bin OS commands were in /usr/src/bin and /usr/src/usr.bin, while the source of locally developed commands was in /usr/local/src, and their binaries in /usr/local/bin. There was no notion of packaging (outside tarballs).

On the other hand, /opt is a directory for installing unbundled packages, each one in its own subdirectory. They are already built whole packages provided by an independent third party software distributor. Unlike /usr/local stuff, these packages follow the directory conventions (or at least they should). For example, someapp would be installed in /opt/someapp, with one of its command being /opt/someapp/bin/foo, its configuration file would be in /etc/opt/someapp/foo.conf, and its log files in /var/opt/someapp/logs/foo.access.

Answer 2

The basic difference is that /usr/local is for software not managed by the system packager, but still following the standard unix deployment rules.

That's why you have /usr/local/bin, /usr/local/sbin /usr/local/include etc...

/opt on the other hand is for software that doesn't follow this and is deployed in a monolithic fashion. This usually includes commercial and/or cross-platform software that is packaged in the "Windows" style.

Answer 3

They are very similar indeed, and the use of one or the other is more a matter of opinion. Linux journal had this point/counterpoint discussion about this exact topic here.

Answer 4

First, I don't think there is a strict answer; different adminstrators will have different opinions, according to their background. Historically, /usr/local came first; it was the convention in Berkley, IIRC. At one point during the development of System V, if I'm not mistaken (this is all a long time ago, and I didn't take notes), there was a decision or a desire to be able to mount /usr read-only, which meant you couldn't add new software to it; that may have been why /opt was invented. As it happens, there was just so much existing software that did write to /usr that that idea never really got off the ground.

My personal preference is /opt, with a separate subdirectory for each product; this makes removing a product a simple case of rm -fr. But if all of your software is installed via a good package manager, it doesn't matter, and if the software you install doesn't strictly obey these conventions, and writes configurations and such somewhere under /usr, it doesn't matter either, although for the opposite reasons.

Answer 5

For me, personally, it's what Bill said in @philfr's link:

On a development system, or a sandbox, having an /opt directory where you can just toss things and see if they work makes a whole lot of sense. I know I'm not going to go through the effort of packaging things myself to try them out. If the app doesn't work out, you can simply rm the /opt/mytestapp directory and that application is history. Packaging may make sense when you're running a large deployment (there are times when I do package apps), but lots of times, it's nice to toss stuff in /opt.

Unfortunately, most make install scripts pushes files into /usr/local instead of just making a symlink there :-/

Answer 6

I have a slightly different take on this issue.
While everything in jlliagre's answer is correct, the practical application for me, when deploying software in a cluster, comes down to default environment variables and default reuse of libs.

Simply put - /usr/local and all its child dirs are in the appropriate env vars such as PATH and MANPATH, and /usr/local/lib{,64} are in ldconfig's (/etc/ld.so.conf.d/).

/opt/ OTOH is not — which is both advantageous when requiring multiple versions or conflicting packages to co-exist in the system, but requires some sort of environment management (e.g., environment-modules or software collections), and disadvantageous in that it would potentially "waste" storage space by duplicating shared libraries, as each installation in /opt can be completely self contained.

For the shared nature of /usr/local to work, it is assumed that e.g. binaries are installed directly to /usr/local/bin (and man pages to appropriate /usr/local/share/man/...) rather than /usr/local/app/{bin,share/man,...} etc.