[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: Final WC format bump before branch

From: Greg Stein <gstein_at_gmail.com>
Date: Tue, 5 Jul 2011 19:13:44 -0400

On Tue, Jul 5, 2011 at 17:47, Hyrum K Wright <hyrum_at_hyrumwright.org> wrote:
>...
> Greg, in r879400 you added the following comment to wc-metadata.sql:
>
> /* Format 99 drops all columns not needed due to previous format upgrades.
>   Before we release 1.7, these statements will be pulled into a format bump
>   and all the tables will be cleaned up. We don't know what that format
>   number will be, however, so we're just marking it as 99 for now.  */
>
> I'm curious as to what has caused you the change your mind.

It seemed like a good idea at the time. After we've been through all
this, it just doesn't seem useful to take time to clear that stuff
out.

>...
> The bulk of the proposed work involved removing stale columns and
> renaming others.  (For various reasons, SQLite allows the addition of
> columns to existing tables, but removing columns involves a
> dump/reload through an intermediate temporary table.)  Dropping extra
> columns saves (an arguably trivial amount of) space, but more
> importantly cleans up a schema that we're going to have to support for
> a long time yet.

Why does the schema need to be "clean"? What benefit will that truly give us?

> It makes sense to do this now, when the only people impacted are devs
> running intermediate versions, rather than the Entire World of
> installed 1.7.x clients.  It's going to happen sometime, let's make it
> impact as few people as possible.

At some point, we'll need to issue a format bump for some reason. Why
not just do it then?

My point is: if we're trying to get this thing stable to produce a
1.7-rc1, then this is one of the *last* things that we should be
doing. It is kind of hard to imagine something potentially more
destabilizing than mucking around with the schema and the statements.
If we forget to switch a ?4 to a ?3, then we introduce a bug. If we
re-arrange some columns, and forget to alter a column fetch for a
'select *' or the bindings for an insert... bug.

At a point where we'd like to stabilize, this doesn't seem like the
right path. And I see zero benefit to us or our end users.

Cheers,
-g
Received on 2011-07-06 01:14:21 CEST

This is an archived mail posted to the Subversion Dev mailing list.