function [mvals]=get_meta_ncml(url);
% GET_META_NCML  Get lon,lat,start,stop from an WHSC time series file 
% assess whether timebase is uniformly spaced
% return a structure of useful stuff for future use
% RPS & ETM 5/2008
nc=netcdf(url);
mvals.lon=nc{'lon'}(1);
mvals.lonmeta=nc.longitude(:);
mvals.lat=nc{'lat'}(1);
mvals.latmeta=nc.latitude(:);
% get the instrument type if available
if isempty(nc.INST_TYPE(:))
    mvals.sensr='unavailable';
else
    mvals.sensr=nc.INST_TYPE(:);
end

% Assume that if BBV, data has a uniform time base 
% the values in time2 are often close but exactly the specified
% Delta_T.  The instrument's precision most likely isn't to the msec, so it
% may be a red hearing to try to use it to determine dt.
ntimes=length(nc('time'));
d1=nc{'time'}(1);

% round time2 values to nearest second to cover for previous sins
% in processing which have led to 1 millisecond errors 
t1=round(nc{'time2'}(1)/1000)*1000;
tend=(nc{'time'}(end)-d1)*3600*24*1000+round(nc{'time2'}(end)/1000)*1000;

% use the attribute for delta_T if available
if (strmatch('DELTA_T',ncnames(att(nc))))
    if isnumeric(nc.DELTA_T(:))
        dt=nc.DELTA_T(:)*1000;  %as msec
    else
        % earliest files had 60.0 minutes in DELTA_T
        if (~isempty(strfind(nc.DELTA_T(:),'min')))
            lx=findstr('.', nc.DELTA_T(:));
            dt=str2num(nc.DELTA_T(1:lx))*60*1000;
        else  % should be a number with seconds assumed
          dt=str2num(nc.DELTA_T(:))*1000;  %as msec
        end
    end
else %use the mean diff of values within a day
    dd=find(abs(diff(nc{'time2'}(:))) > 8630000);
    dt=mean(diff(nc{'time2'}(dd(5)+1:dd(6))))
end
% Express start/stop times in Matlab datenum

mvals.jd_start=(d1 - 1721059)+t1/3600/24/1000;
mvals.jd_start=(d1 - 1721059)+tend/3600/24/1000;
% Calculate start/stop times in Matlab datenum with units=milliseconds
mvals.mjd_start=(d1 - 1721059)*3600*24*1000 + t1;
%mvals.mjd_stop =(d1 - 1791059)*3600*24*1000 + tend;  % typo
mvals.mjd_stop =(d1 - 1721059)*3600*24*1000 + tend;

tot1=dt*(ntimes-1);
tot2=tend-t1;
% accept as uniform, even if a sceond off
if (tot1==tot2) | (abs(tot2-tot1) <= 1000)
    mvals.timebase_uniform=1;
    mvals.dt=dt/1000;  % dt in seconds
    mvals.dtms=dt;  % dt in milliseconds
elseif abs(tot2-tot1) <= 10000   % within 10 seconds
    disp('within 10 seconds- do something here') 
else    
    mvals.timebase_uniform=0;
end

% use WATER_DEPTH or water_depth attribute
if isempty(nc.WATER_DEPTH(:))
    if isnumeric(nc.water_depth(:))
        mvals.wdep=nc.WATER_DEPTH(:);
    else
        mvals.wdep='missing';
    end
else
    mvals.wdep=nc.WATER_DEPTH(:);
end
mvals.meas_dep=nc{'depth'}(:);
mvals.varnames=ncnames(var(nc));
for ik=1:length(mvals.varnames)
  eval(['mvals.long_names{ik} = nc{''' mvals.varnames{ik} '''}.long_name(:);'])
end
close(nc);