Stefan Sperling <[email protected]> writes: > On Fri, Jul 08, 2011 at 10:44:27AM +0100, Julian Foad wrote: >> There's a comment in libsvn_ra_[neon|serf]/options.c saying "we should >> probably remove it before 1.7 goes final". Should we remove or keep it? >> >> #define SVN_IGNORE_V2_ENV_VAR "SVN_I_LIKE_LATENCY_SO_IGNORE_HTTPV2" >> >> #ifdef SVN_DEBUG >> /* ### This section is throw in here for development use. It >> ### allows devs the chance to force the client to speak v1, >> ### even if the server is capable of speaking v2. We should >> ### probably remove it before 1.7 goes final. */ >> char *ignore_v2_env_var = getenv(SVN_IGNORE_V2_ENV_VAR); >> >> if (! (ignore_v2_env_var >> && apr_strnatcasecmp(ignore_v2_env_var, "yes") == 0)) >> ras->me_resource = apr_pstrdup(ras->pool, val); >> #else >> ras->me_resource = apr_pstrdup(ras->pool, val); >> #endif > > It is easier to set an environment variable than installing an older > client that only speaks the old protocol. > > So it might be useful to keep this to facilitate problem diagnosis > once httpv2 clients are in the field. In which case the comment > should be adjusted, of course, and we should move this out of SVN_DEBUG.
If we are going to keep it then we should make it a config option like http-compression (and avoid a name with 'httpv2' to make it clear that this is a Subversion protocol). -- uberSVN: Apache Subversion Made Easy http://www.uberSVN.com
