Attending: Jens (c+m), RobC, Teng, Sam, MattD, PeteC, Brian, Winnie, Steve 1. Report from last week's things Sam reported lots of machine learning activities (attempts?) In SadM, lots of talks about data caching and data movement, but sometimes not so practical, not addressing how much cache is actually needed or how workflows might change to benefit from the caching. EOS is doing the erasure code thing but not using parity bits (blocks) for recalling data which would have been more efficient despite the very very slightly more complicated "computation." Also the usual talks about non-X.509 authorisation; this time, again, both the scitoken and the macaroons were present. Plenaries on data placement and SKA seems to not have funding for regional data analysis. ICECUBE apparently looking at Rucio. Unrelated to CHEP, Pete had had a discussion with John Illingworth from FNAL about Rucio. BOFs - there was a DOMA one which spent more time discussing the relative merits of xroot and http; and one on Rucio. Perhaps scope to look again at non-TCP transfers; lots of past work on UDP and derivatives thereof. Is it worth doing this work again? only if something's changed. Proposal to use blockchains: one of the (many) annoying things about bitcoins is the proof of work which is essentially useless in the Grand Scheme of Things; it serves no purpose from a science point of view other than to convert electricity into heat. Several people have thought about doing more useful computational stuff as PoW, but how are they to be rewarded for it? Useless though it is, the value of the PoW in bitcoin is the reward for "mining" it. Could one provide a proof of data stored (and checked by sampling bits that the data were still held)? ... clearly needs more thought; is it something we (GridPP) would be interested in? Tapes: ATLAS carousel; dataflow from tape, like whether ATLAS' data is split across tapes. 2. This quarter's things Not much progress on the GridPP6 forward view (in fact, none at all.) Link reshared (privately). Makes sense to make this a reportable task in the QR. Other topics could be the CHEP presentation(?) or GridPP in August. 3. AOB Teng: (18/07/2018 10:09:50) https://indico.cern.ch/event/587955/contributions/2936865/ https://indico.cern.ch/event/587955/contributions/2936816/attachments/1679283/2697239/CHEP-2018-07-09-cheidecker-v2.pdf https://indico.cern.ch/event/587955/contributions/2936833/ Brian Davies @RAL-LCG2: (10:16 AM) any further discussion on tape , especially on ATLAS carousel model? Steve Jones: (10:30 AM) It would be good if we could process data instead of meaninglessly searching cyberspace for random files that have a lot of zeroes in their MD5sums! Brian Davies @RAL-LCG2: (10:32 AM) rucio discusiion will bo into the data manaagement section I aded.