We report on GHRS observations which reveal conspicuous differences in
the absorption spectra of two nearby stars, close to each other. The
star HD102065 lies behind a cloud in Chamaeleon with unusually strong
mid-LR emission, indicating a large abundance of very small dust parti
cles. Along this line of sight, 5% of the gas (about 6 10(19) cm(-2))
is at large velocities (up to -50 km s(-1)) compared to the main absor
ption component at v(lsr) similar to 0 km s(-1). The high velocity gas
is very excited and has an unusually large silicon abundance. The oth
er star HD96675 lies behind a cloud with standard mid-IR emission. Alo
ng this line of sight, high velocity gas is also detected, but to some
what smaller offset velocities and with a much lower excitation. In pa
rticular, the SiII lines are not observed. From the excitation of Si and C+ in the direction of HD102065, we infer that the high velocity
gas has a temperature higher than several 100 K and an electron densit
y of at least 10 cm(-3). These results, together with the lack of an i
onizing star in the neigborhood, suggest that a large amount of kineti
c energy is being deposited in this gas. The collision of an infalling
cloud and a local cloud is a plausible source of energy. The peculiar
ity of the dust size distribution inferred from the IRAS data is likel
y to be related to the processes which dissipate the kinetic energy an
d heat the gas. A shock seems to be required to produce the excitation
and ionization degrees. But the carbon ionization ratio combined with
the electron density inferred from the silicon and carbon excitation
implies that carbon is not in ionization equilibrium and should recomb
ine extremely quickly. Future higher resolution observations might hel
p solving this incoherency.