We present Hubble Space Telescope UV spectrophotometry of three extremely metal-deficient blue compact dwarf (BCD) galaxies, SBS 0335-052 (Z similar to Z(.)/40), Tol 65 (Z similar to Z(.)/24), and T1214-277 (Z similar to Z(.)/23). Broad damped Ly alpha absorption is seen in the first two BCDs. For SBS 0335-052, the H I column density derived by fitting the Ly alpha absorption profile is N(H I) = (7.0 +/- 0.5) x 10(21) cm(-2), which is the highest H I column density derived thus far for a BCD and which is similar to 2 times larger than that in I Zw 18. As for Tol 65, N(H I) = (2.5 +/- 1.0) x 10(21) cm(-2), which is also in the high range of H I column densities derived for BCDs and which is only -1.4 times smaller than that in I Zw 18. The interstellar absorption line O I lambda 1302 has also been detected in both galaxies. Comparison with high-resolution quasar spectra implies that in SBS 0335-052, the O I lambda 1302 line, along with other heavy element interstellar absorption lines such as Si II lambda 1304 and S II lambda 1251, lambda 1254, and lambda 1259, are not saturated, which allows us to derive abundances. Assuming that these lines originate in the H I gas, we derive abundances of oxygen, silicon, and sulfur, respectively, as 37,000, 4000, and 116 times lower than the solar values. The oxygen abundance is a whole 37 times lower than in the neutral gas of I Zw 18. However, these highly discrepant deficiency factors of different elements suggest that the absorption lines are produced, not in the H I, but in the H II gas. Adopting that hypothesis, the derived abundances from the UV absorption lines are then consistent with that derived from the optical emission lines (Z similar to Z./40). The conclusion that the heavy element absorption lines originate in the H II region is supported by the detection of several systems of blueshifted S II lambda 1259, Si II lambda 1260, O I lambda 1302, Si II lambda 1304, and C II lambda 1335 absorption lines originating in fast-moving clouds with velocities up to similar to 1500 km s(-1) and also by the presence of heavy element absorption lines with excited lower levels. If this conclusion holds, then the H I cloud in SBS 0335-052 is truly primordial, unpolluted by heavy elements. This would alleviate the need for postulating previous enrichment by widespread Population III stars and poses the problem of how such a cloud could last for a Hubble time without making stars before the present epoch. Because the O I lambda 1302 is likely to be saturated in Tol 65, we could derive only a lower limit for the oxygen abundance, similar to 6000 times lower than the solar value. Contrary to the situation in the two previous BCDs, a strong Ly alpha line is seen in the spectrum of T1214-277, which makes it the lowest metallicity BCD with detected Ly alpha emission. Its equivalent width of 70 Angstrom is the largest found in star-forming galaxies. The Ly alpha emission is not redshifted with respect to the H II gas velocity, so that the escape of Ly alpha photons in T1214-277 is not dictated by the velocity structure of the H I gas but probably by its porosity. The absence of Ly alpha emission in SBS 0335-052 and Tol 65 is probably caused by a combination of dust extinction, redistribution of the Ly alpha photons by multiple scattering over the whole area of the H I cloud, and the geometry of the cloud. The spectrum of SBS 0335-052 shows stellar Si IV lambda 1394 and lambda 1403 lines with P Cygni profiles, which suggests the presence of numerous hot supergiant stars. A strong stellar N V lambda 1240 line with a P Cygni profile is also seen in the spectrum of T1214-277, which suggests that a large population of massive (M greater than or similar to 60 M.) stars is present. The stellar wind terminal velocity is very low in SBS 0335-052, being only similar to 500 km s(-1). It is higher, similar to 2000 km s(-1) in T1214-277, in the low range of terminal velocities found for other BCDs. These BCDs are the two most metal-deficient galaxies known with P Cygni profiles. The presence of such profiles raises the question of how to set up a wind in stellar atmospheres devoid of heavy elements as in SBS 0335-052.