Wowledge communities of ti kinds have social ad material practices for deciding what is known and who is to be -h ti paper, we address a spec%c kind of knowledge work environrnenti planning, and a partictdar form of coflaboratio~tie sharing ofmeasurement data sets.We xe interested in how trust is create~how trustab%ty is assessed in the arrn's-length co~aboration of sharing data sex and how changes in technolo~intemct tith those practices of ~L We look at seveti elements of scientic pmctice that ficfitate this sharing -communities of practice, boundary objects, and assemblages -and discuss the impficntions for CSCW, digiti fibraries, and other tiormation-sharing applications.Ke~ords fiowledge u'or~enviromnenti plamting, boundary objects, assemblages, communities of practice, data sets, digiti kiraries, UC Berkeley Dim Library lNTRODUC~ON Technology desi~ed to support work has to be consonant ~~titbacti work practices.CSCW is e~ficitiy concerned ujth the socifly-ormd practices of users, as we~as with technologi~support for those practices.hluch CSCW research and design has been directd at stnti, stable, heterogeneous groups with on-going org-oti andor task connections [e.g.1S, 164, et d].Even when membership in the group is fily lmge, it is, or is assumed to be, tited andor know e.g.people working on a projec~within an ormtio~or as part of a professional community.However, much cooperative work is performed by large, heterogeneous, sfig, transient groups,distributed in time and space, with permeabIe bounties.Permimion to make di~bl or hard copies of all or part of this \vork for paonal or classroom use is granted without fee provided that copies are not made or dism%utedfor profit or commercial ad%=tage and that copiesbear this notice and the fill citation on the firstpaga To copy otim~is~to Tepublish,to post on senrers or to rdlsm%ute to ~its, requires prior specific perrnision an&'ora f= CSC\V 98 Seattle IVxtington USA Cop~ti@t ACLI 199S 1-5S1134094/9S/1 I...00OO.