Message6110

Author malte
Recipients florian, gabi, jendrik, malte, salome, silvan
Date 2017-01-13.01:48:50
Content
I don't really understand the code either. Silvan, Salomé and Manuel have been
involved in it more recently, and perhaps they have an idea if the change makes
sense, although this is probably a part of the code none of us really
understands. :-)

This cannot affect our optimal configurations since these don't use reasonable
orders, so I guess the worst thing that can happen if we break this is that we
make LAMA-style configurations a bit less efficient, which is something we can
test experimentally. (Potential bugs for optimal configurations are a larger
worry because we might we break admissibility by deriving wrong orders, and
ostensibly optimal configurations performing non-optimally is a worse sin than
bad performance.)

Hence, I'm also fine with merging this after an experiment. I suggest testing
with lama-first and with seq-sat-lama-2011. Because the latter is anytime, they
would need different reports, I think. (For the anytime configurations, many of
our attributes don't make sense.)

More long-term, we definitely need to clean up the landmark code. But I think
this is more something for a "task force" than for the review queue. If someone
is interested in this, I suggest we discuss the topic further once I'm back in
office.
History
Date User Action Args
2017-01-13 01:48:50maltesetmessageid: <1484268530.1.0.703577447161.issue383@unibas.ch>
2017-01-13 01:48:50maltesetrecipients: + malte, gabi, jendrik, silvan, florian, salome
2017-01-13 01:48:50maltelinkissue383 messages
2017-01-13 01:48:50maltecreate